/h/ - Hentai » Thread #7382540

[1223 / 318]

/hdg/ - Hentai Diffusion General #324

Anonymous Thu 18 May 2023 20:00:59 No.7382540 View View Reply Original Report

Quoted By: >>7382574 >>7382669 >>7382768 >>7383628 >>7384379 >>7394315

Good Ol' Edition

Old: >>7379025

>RESOURCES
Wiki: https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki
LoRAs: https://gitgud.io/gayshit/makesomefuckingporn | https://github.com/derrian-distro/LoRA_Easy_Training_Scripts
Tags: https://danbooru.donmai.us/wiki_pages/tag_groups
ControlNet: https://rentry.org/dummycontrolnet
Animation: https://rentry.org/AnimAnon
Wildcards: https://rentry.org/NAIwildcards
4chanX Catbox userscript: https://gist.github.com/catboxanon/ca46eb79ce55e3216aecab49d5c7a3fb
OP Template/Logos: https://pastebin.com/0atr0fEc | https://files.catbox.moe/xk99gy.png | https://files.catbox.moe/35bdt5.gif

Anonymous

View Same Google ImgOps iqdb SauceNAO 515470970.png, 1MiB, 1024x1280

Anonymous Thu 18 May 2023 20:02:19 No.7382547 Report

Quoted By: >>7382569

old but gold

Anonymous

Anonymous Thu 18 May 2023 20:02:23 No.7382548 Report

Quoted By:

placebo edition

Anonymous

View Same Google ImgOps iqdb SauceNAO kek2.png, 642KiB, 873x618

Anonymous Thu 18 May 2023 20:03:41 No.7382553 Report

Quoted By: >>7382555 >>7382558 >>7382581

this for the next thread op

Anonymous

Anonymous Thu 18 May 2023 20:04:02 No.7382555 Report

Quoted By: >>7382581

>>7382553
that's not an /h/ image you can't even see tits retard

Anonymous

Anonymous Thu 18 May 2023 20:04:50 No.7382558 Report

Quoted By: >>7382562 >>7391449

>>7382553
Am I supposed to know what this is?

Anonymous

Anonymous Thu 18 May 2023 20:06:41 No.7382562 Report

Quoted By: >>7391449

>>7382558
probably some tiktok meme

Anonymous

View Same Google ImgOps iqdb SauceNAO 9996490989765790-1547945326-kill (...).png, 1MiB, 1024x1536

Anonymous Thu 18 May 2023 20:10:55 No.7382567 Report

Quoted By: >>7382571 >>7382573 >>7382575 >>7383210

Anonymous

View Same Google ImgOps iqdb SauceNAO (you) should inpaint.jpg, 191KiB, 1024x1280

Anonymous Thu 18 May 2023 20:11:11 No.7382569 Report

Quoted By: >>7382576

>>7382547
are you still doing the needful and devoting yourself to inpainting and controlnet?

Anonymous

View Same Google ImgOps iqdb SauceNAO 9996490989765779-1057863109-kill (...).png, 2MiB, 1600x2000

Anonymous Thu 18 May 2023 20:11:44 No.7382571 Report

Quoted By:

>>7382567

Anonymous

View Same Google ImgOps iqdb SauceNAO 23022-1166416034-Yugi,__nipples, (...).png, 1MiB, 1024x1536

Anonymous Thu 18 May 2023 20:11:58 No.7382572 Report

Quoted By:

screwing with styles gives some hilarious fake screenshots at times

Anonymous

Anonymous Thu 18 May 2023 20:12:17 No.7382573 Report

Quoted By:

>>7382567
dayum

Anonymous

Anonymous Thu 18 May 2023 20:12:34 No.7382574 Report

Quoted By: >>7382592

>>7382540
>OP still not updated

Anonymous

Anonymous Thu 18 May 2023 20:13:42 No.7382575 Report

Quoted By:

>>7382567
Noice

Anonymous

Anonymous Thu 18 May 2023 20:13:46 No.7382576 Report

Quoted By:

>>7382569
always

Anonymous

View Same Google ImgOps iqdb SauceNAO 00119-3133574008.png, 3MiB, 1856x1152

Anonymous Thu 18 May 2023 20:17:04 No.7382578 Report

Quoted By: >>7382579 >>7382580 >>7383465

Anonymous

Anonymous Thu 18 May 2023 20:17:47 No.7382579 Report

Quoted By:

>>7382578
chaos prompt

Anonymous

Anonymous Thu 18 May 2023 20:17:55 No.7382580 Report

Quoted By: >>7382591

>>7382578
what are they rioting over

Anonymous

View Same Google ImgOps iqdb SauceNAO samantha-goldstriker.png, 654KiB, 873x618

Anonymous Thu 18 May 2023 20:18:26 No.7382581 Report

Quoted By:

>>7382553
>>7382555

Anonymous

View Same Google ImgOps iqdb SauceNAO 00007-676712000.png, 339KiB, 512x512

Anonymous Thu 18 May 2023 20:19:00 No.7382582 Report

Quoted By: >>7382585 >>7382586 >>7382587 >>7382604

noob here.
I'm trying to make pictures of a particular character but it's not recognizing the character, I've put her name and the name of the property she's from in the tags as well as a couple details about her. but it's giving me a totally different girl. the tags are pretty simple but do you think this would be a tag issue or could it be an issue with the model I picked. I checked some of the images in the last thread and basically picked a model at random. It's Based64_v3. The character i want to make images of is Sessyoin Kiara from Fate if anyone is wondering.

Anonymous

Anonymous Thu 18 May 2023 20:20:10 No.7382585 Report

Quoted By: >>7382617

>>7382582
apply vae and get a lora

Anonymous

Anonymous Thu 18 May 2023 20:21:36 No.7382586 Report

Quoted By: >>7382617

>>7382582
vae to fix colors and look for her lora on civit

Anonymous

Anonymous Thu 18 May 2023 20:22:07 No.7382587 Report

Quoted By:

>>7382582
average fate enjoyer

Anonymous

Anonymous Thu 18 May 2023 20:24:23 No.7382591 Report

Quoted By:

>>7382580
penis denial

Anonymous

Anonymous Thu 18 May 2023 20:24:29 No.7382592 Report

Quoted By: >>7382597 >>7382605

>>7382574
>updated to what

Anonymous

View Same Google ImgOps iqdb SauceNAO 108225550_p2_master1200.jpg, 263KiB, 512x784

Anonymous Thu 18 May 2023 20:24:47 No.7382594 Report

Quoted By: >>7382604 >>7382605

>>7382570
>its just fans of a single series in my case,
By serious do you mean a certain obscure anime or your own OCs?

Anonymous

View Same Google ImgOps iqdb SauceNAO 00210-46561232.png, 2MiB, 1024x1536

Anonymous Thu 18 May 2023 20:25:24 No.7382595 Report

Quoted By:

Anonymous

View Same Google ImgOps iqdb SauceNAO 06815-3851538426-nsfw, masterpie (...).png, 1008KiB, 768x1152

Anonymous Thu 18 May 2023 20:26:02 No.7382596 Report

Quoted By: >>7382600 >>7382607

new thread new cow (for breeding)

Anonymous

Anonymous Thu 18 May 2023 20:27:01 No.7382597 Report

Quoted By: >>7382603

>>7382592
updated to restore the outdated

Anonymous

Anonymous Thu 18 May 2023 20:27:16 No.7382600 Report

Quoted By:

>>7382596
>new cow
Clearly this image is from 2012!

Anonymous

Anonymous Thu 18 May 2023 20:28:17 No.7382603 Report

Quoted By:

>>7382597
kek

Anonymous

View Same Google ImgOps iqdb SauceNAO e9a4316ee37040ee7e9ba761b7d43fbc_2.jpg, 356KiB, 864x1328

Anonymous Thu 18 May 2023 20:28:20 No.7382604 Report

Quoted By: >>7382608 >>7382610 >>7382617

>>7382594
*Series

>>7382582
What you're asking is like asking How do I make them a millennium falcon Lego ship out of a Lego piece set that only has Barbie pieces. You can't prompt what isn't there.

>LoRAs: gitgud.io/gayshit/makesomefuckingporn | github.com/derrian-distro/LoRA_Easy_Training_Scripts
https://youtu.be/70H03cv57-o

I highly recommend you find a Google colab Lora trainer that has a built-in image tagger. Good luck. You're gonna need patience for what you're asking for

Anonymous

Anonymous Thu 18 May 2023 20:28:43 No.7382605 Report

Quoted By: >>7382611 >>7382615

>>7382594
obscure. Make a lora for the cgi monster, we need the real god to take the stage.

>>7382592
someone asked what to put in next op and people replied, you didnt even include install guides

Anonymous

View Same Google ImgOps iqdb SauceNAO Snapchat-709276277.jpg, 7KiB, 125x118

Anonymous Thu 18 May 2023 20:28:59 No.7382607 Report

Quoted By:

>>7382596
Sir you just doxxed yourself

Anonymous

Anonymous Thu 18 May 2023 20:29:36 No.7382608 Report

Quoted By: >>7382615

>>7382604
>youtube guide

Anonymous

View Same Google ImgOps iqdb SauceNAO 01180-247836925.png, 322KiB, 512x768

Anonymous Thu 18 May 2023 20:29:40 No.7382609 Report

Quoted By:

Anonymous

Anonymous Thu 18 May 2023 20:29:44 No.7382610 Report

Quoted By: >>7382615

>>7382604
>the accursed youtube link
why would you do this

Anonymous

Anonymous Thu 18 May 2023 20:30:26 No.7382611 Report

Quoted By: >>7382618 >>7382629

>>7382605
>someone asked what to put in next op and people replied, you didnt even include install guides
Nobody said anything useful, so yeah.
The install guides are under the wiki. Updated.

Anonymous

View Same Google ImgOps iqdb SauceNAO g_1684156870754700.png, 4MiB, 1280x1920

Anonymous Thu 18 May 2023 20:31:54 No.7382615 Report

Quoted By: >>7382619 >>7382627

>>7382605
>cgi monster,
Which CGI monster?

>>7382610
>>7382608
Because the newb needs a retard-proof guide and the guides we usually use (The ones that keeping linked in the OPs) are written by people who have never had to teach anyone anything in their life. Some have flat out incorrect information so I'm hesitant to link those most of the time.

Anonymous

Anonymous Thu 18 May 2023 20:32:29 No.7382617 Report

Quoted By:

>>7382604
>>7382585
>>7382586
thanks frens. excuse my autism.

Anonymous

Anonymous Thu 18 May 2023 20:33:16 No.7382618 Report

Quoted By: >>7382630

>>7382611
>nobody said anything useful
stop pushing your retardation faggot
>quick install guides inside some wiki link
you could link the relevant wiki page dumbass

Anonymous

Anonymous Thu 18 May 2023 20:33:16 No.7382619 Report

Quoted By: >>7382624

>>7382615
>choses to link guide that is misinformation at best
BASED RETARD

Anonymous

View Same Google ImgOps iqdb SauceNAO wipe.png, 1MiB, 768x1152

Anonymous Thu 18 May 2023 20:33:33 No.7382620 Report

Quoted By: >>7384309

Anonymous

View Same Google ImgOps iqdb SauceNAO ic_1683753548746578.png, 42KiB, 185x156

Anonymous Thu 18 May 2023 20:34:50 No.7382624 Report

Quoted By: >>7382632

>>7382619
The YouTube video or the gitgud page?

Anonymous

Anonymous Thu 18 May 2023 20:35:46 No.7382627 Report

Quoted By: >>7382633

>>7382615
>CGI
did you make a kamisama lora without watching the show? cringe
>guide
you linked the worst one thats 100% misinformation genius

Anonymous

Anonymous Thu 18 May 2023 20:36:53 No.7382629 Report

Quoted By: >>7382639 >>7382644

>>7382611
>the only relevant lora training guide still not here
>apparently not useful
holy shit why are you so fucking retarded

Anonymous

Anonymous Thu 18 May 2023 20:36:53 No.7382630 Report

Quoted By: >>7382644

>>7382618
>stop pushing your retardation faggot
The only things that were said were "put in outdated shit; put in magical good guide that doesn't exist; link dead boards".
You could also bake yourself, but you didn't! I swear this is some form of Cunningham's Law in play.
>you could link the relevant wiki page dumbass
Sure, why not, if people are too retarded to actually look into it themselves.

Anonymous

View Same Google ImgOps iqdb SauceNAO 00214-3497177391.png, 2MiB, 1024x1536

Anonymous Thu 18 May 2023 20:37:17 No.7382631 Report

Quoted By:

Anonymous

Anonymous Thu 18 May 2023 20:37:40 No.7382632 Report

Quoted By: >>7382641

>>7382624
The video guide has had absolute disastrous consequences for lora since it makes retards load in 10 images at 100 repeats and upload their gigafried shit to every sharing site

Anonymous

Anonymous Thu 18 May 2023 20:37:55 No.7382633 Report

Quoted By: >>7382647

>>7382627
How is it misinformation?

Anonymous

View Same Google ImgOps iqdb SauceNAO 00076-751153848-nsfw, scan, (((t (...).png, 2MiB, 1024x1536

Anonymous Thu 18 May 2023 20:39:25 No.7382638 Report

Quoted By: >>7383433

Anonymous

Anonymous Thu 18 May 2023 20:39:34 No.7382639 Report

Quoted By: >>7382652 >>7383009

>>7382629
If you mean the one I think you do, then someone take the 15 minutes to reformat it and make it an actual guide (a dictionary is not a guide) and nobody would be against putting it in, desu.

Anonymous

View Same Google ImgOps iqdb SauceNAO 02407-2087788976-alural(kamikats (...).png, 483KiB, 512x784

Anonymous Thu 18 May 2023 20:40:40 No.7382641 Report

Quoted By: >>7382647

>>7382632
Care to explain why that's necessarily bad? If that's happening with (and I'm assuming it's only with one epoch) only a thousand steps and it still being overbaked then they're learning rate is probably way too high. The Alural LoRA I made the other day has 36 images with 100 repeats per image and did not over bake because I had a reasonable learning rate. The type of schedule are used (constant) probably had a great effect too.

Anonymous

Anonymous Thu 18 May 2023 20:40:59 No.7382644 Report

Quoted By: >>7382646 >>7382648 >>7382660

>>7382630
>link dead boards
how are they dead, /g/ has more posts than this place you fucking schizoid. stop living in your own world
>guide
see >>7382629 and >>7379783 >>7379200 last thread
>people are too retarded
Maybe if you used your whatever's left of your brain you would notice and perhaps realize why every single thread has guides linked directly in the OP and not in some decrepit wiki buried under 30 sections.
I hope you realize you're autistic, could try and se if your gov will give you neetbux cause it seems severe.

Anonymous

Anonymous Thu 18 May 2023 20:41:29 No.7382646 Report

Quoted By: >>7382652

>>7382644
DIY

Anonymous

Anonymous Thu 18 May 2023 20:42:09 No.7382647 Report

Quoted By: >>7382650

>>7382633
100 repeats

>>7382641
cant tell if new or bait, lurk more

Anonymous

Anonymous Thu 18 May 2023 20:42:58 No.7382648 Report

Quoted By: >>7382649 >>7382652

>>7382644
>>>7379783
>DAdaptation needs specific arguments to work. Scheduler must be set to constant
BASED

Anonymous

Anonymous Thu 18 May 2023 20:45:43 No.7382649 Report

Quoted By: >>7382651

>>7382648
if by that you mean "wrong"

Anonymous

Anonymous Thu 18 May 2023 20:45:58 No.7382650 Report

Quoted By: >>7382654

>>7382647
The final number of total repeats and the learning rate are more important. YOU lurk more.

Anonymous

Anonymous Thu 18 May 2023 20:46:33 No.7382651 Report

Quoted By:

>>7382649
Yeah, you need tone indicators or something?

Anonymous

Anonymous Thu 18 May 2023 20:47:34 No.7382652 Report

Quoted By: >>7382653 >>7382658

>>7382646
Then dont bake a thread if you cant be bothered to do it properly, monkey because now no one can do it till the next one.

>>7382639
too many words for zoom zoom? if someone gets filtered by knowledge then I dunno what to tell you. But yeah its not exactly beginner friendly nor is meant to be - people were discussing making a new one for that purpose. Its just a very good information compendium

>>7382648
Not sure if genuine praise or what but the constant thing for dadapt is nothing new. No clue about arguments, I'd assume easyscripts handles it auto?

Anonymous

Anonymous Thu 18 May 2023 20:48:15 No.7382653 Report

Quoted By: >>7382659

>>7382652
constant looks like fucking shit on anything including dadaptation, you should use cosine or cosine with restarts

Anonymous

Anonymous Thu 18 May 2023 20:48:48 No.7382654 Report

Quoted By: >>7382682

>>7382650
>what is shuffling
>why people advocate for less repeats and more epochs
you can stop lurking when you figure these things out

Anonymous

Anonymous Thu 18 May 2023 20:49:18 No.7382656 Report

Quoted By:

How do I make ffmpeg JPGs not look like shit? I train on colab and PNGs are just way too large to upload 1000 of them

Anonymous

Anonymous Thu 18 May 2023 20:49:54 No.7382658 Report

Quoted By: >>7382665

>>7382652
Sorry, but ultimately until you stop being a little scared bitch, it's gonna be me or whoever else that bakes that decide what's in the OP. DIY.
And no, constant for dadapt is made up.

Anonymous

Anonymous Thu 18 May 2023 20:50:27 No.7382659 Report

Quoted By: >>7382663 >>7382664

>>7382653
dadaptation adjusts the learning rate, it wouldnt work properly with cosine because cosnie keeps decaying, its an obvious conflict dumbass.
>looks fucking shit
thats just dadaptation being trash, I've seen people get constant to work but yeah I wouldnt recommend it

Anonymous

Anonymous Thu 18 May 2023 20:51:29 No.7382660 Report

Quoted By: >>7382662 >>7382665

>>7382644
>/g/ has more posts than this place
Ironically the /h/ thread has been far more helpful whenever I've had tech questions compared to /g/ where the posts are just drowned out by people doing a 20 image dump of their 1girl, 3d prompt

Anonymous

Anonymous Thu 18 May 2023 20:52:19 No.7382662 Report

Quoted By:

>>7382660
/g/ is more of a culture war board than a technology board.

Anonymous

Anonymous Thu 18 May 2023 20:52:50 No.7382663 Report

Quoted By: >>7382666 >>7382671

>>7382659
NTA but how about you find the GitHub README for D-Adaptation and read what it says about learning rate schedules?

Anonymous

Anonymous Thu 18 May 2023 20:53:21 No.7382664 Report

Quoted By: >>7382666

>>7382659
while i agree that dadap is useless, this isnt theory ive actually tested schedulers and constant is fuckignshit

Anonymous

Anonymous Thu 18 May 2023 20:54:15 No.7382665 Report

Quoted By: >>7382668

>>7382660
I agree its a shit thread but that doesn't make it "dead" or a reason to unlist it from OP.

>>7382658
>Sorry, but ultimately until you stop being a little scared bitch, it's gonna be me or whoever else that bakes that decide what's in the OP. DIY.
>I BAKED FIRST SO I DECIDE WHATS IN THE OP11!1!11
fuck off back to where you came from, this isn't how things work here

Anonymous

Anonymous Thu 18 May 2023 20:54:22 No.7382666 Report

Quoted By:

>>7382663
>>7382664
nooo he's gonna call you retards now :(

Anonymous

Anonymous Thu 18 May 2023 20:55:41 No.7382668 Report

Quoted By: >>7382676

>>7382665
>fuck off back to where you came from, this isn't how things work here
Clearly it is, because you've been assblasted for good two weeks now and hasn't once baked your good OP, lmao!
Maybe next time. Should prepare it now so nobody does it before you.

Anonymous

Anonymous Thu 18 May 2023 20:55:44 No.7382669 Report

Quoted By: >>7382725

>>7382540
Where do I go if I want to learn how to make those photorealistic images I've seen on Pixiv? Is this the correct thread or is there one on another board? This looks more anime naturally.

Anonymous

Anonymous Thu 18 May 2023 20:56:55 No.7382671 Report

Quoted By: >>7382674 >>7382675 >>7382698

>>7382663
who cares what the readme says? its not like the author extensively tested it on all the schedulers to find out which one works best

Anonymous

Anonymous Thu 18 May 2023 20:57:35 No.7382674 Report

Quoted By:

>>7382671
BASED

Anonymous

Anonymous Thu 18 May 2023 20:57:55 No.7382675 Report

Quoted By: >>7382678

>>7382671
ok now you're just trolling

Anonymous

Anonymous Thu 18 May 2023 20:58:17 No.7382676 Report

Quoted By: >>7382679

>>7382668
>everyone complaining is the same person
whatever. keep vandalizing the thread and ignoring people, this place's a goner

Anonymous

Anonymous Thu 18 May 2023 20:59:26 No.7382677 Report

Quoted By:

the waving tag is fucking cursed

Anonymous

Anonymous Thu 18 May 2023 20:59:40 No.7382678 Report

Quoted By:

>>7382675
so you're telling me kohya knows everything there is to know about loras and that if he says something it certainly is true? I can recall this not being the case way too many times.

Anonymous

Anonymous Thu 18 May 2023 21:01:15 No.7382679 Report

Quoted By: >>7382685

>>7382676
DIY applies to you too, matey. Bake a template and if people like it, they'll use it. Same as with this one.
I think I should actually vandalize the next OP so Cunningham's goes into full effect and we get a wave of people suddenly knowing what to put in the OP. Heh.

Anonymous

Anonymous Thu 18 May 2023 21:01:46 No.7382680 Report

Quoted By:

>I made something so I know everything there is to know
damn, game devs keep winning all the tourneys huh

Anonymous

View Same Google ImgOps iqdb SauceNAO gif_1684248226190294.gif, 3MiB, 291x400

Anonymous Thu 18 May 2023 21:02:34 No.7382682 Report

Quoted By: >>7382686

>>7382654
>Fails to explain why high repeats is inherently bad when it just werks as long as you're not too retarded and too lazy to adjust your learning rates

Anonymous

Anonymous Thu 18 May 2023 21:02:54 No.7382683 Report

Quoted By: >>7382684 >>7382691

>i made a random assertion based on nothing so I know the truth
based

Anonymous

Anonymous Thu 18 May 2023 21:04:59 No.7382684 Report

Quoted By: >>7382690

>>7382683
>desu it is

Anonymous

Anonymous Thu 18 May 2023 21:05:29 No.7382685 Report

Quoted By: >>7382690

>>7382679
>Bake a template and if people like it, they'll use it
OPs dont operate on a case by case like or dislike by the baker, but its not my job to each autistic newfags about site culture so go do whatever, this place's almost as bad as /g/ now.

Anonymous

Anonymous Thu 18 May 2023 21:06:30 No.7382686 Report

Quoted By: >>7382717

>>7382682
>wasting image limit on garbage
you were told all you need to go search the archive, not my problem. if you're going to cope limit it to text

Anonymous

Anonymous Thu 18 May 2023 21:07:08 No.7382687 Report

Quoted By: >>7382692 >>7382693

Does a higher batch size always result in longer gen times? Seems to increase linearly but I thought having high VRAM would allow me to gen more at the same speed. Don't see the reason to use high batch size then if batch count would just do the same thing.
Any 3060 bro here that could tell me good settings would be greatly appreciated, not the resolution, steps, cfg, hires stuff just count/size and other settings.

Anonymous

Anonymous Thu 18 May 2023 21:09:10 No.7382690 Report

Quoted By: >>7382694 >>7382699

>>7382685
>OPs dont operate on a case by case like or dislike by the baker
Oh yeah right, I forgot to call the /hdg/ High Council. Fuck off. Make a better template and it will get accepted and used just like this one has.
Speaking of, anyone know of a good model recommend site/repo to link? It'd mostly have to be a collection of the good AOM2 mixes, desu. I want to put something in there but linking shit like Protogen to normalfags is not good.
>>7382684
Yeah, true.

Anonymous

Anonymous Thu 18 May 2023 21:09:27 No.7382691 Report

Quoted By: >>7382693

>>7382683
>random assumption based on nothing
>guide based off testing for months is "nothing"
Surely dadaptation's author has baked many loras with it and not just focused on making a working scheduler. Surely.

Anonymous

Anonymous Thu 18 May 2023 21:09:36 No.7382692 Report

Quoted By:

>>7382687
I just do 1 for both but I also tend to continuously work on one image at a time rather than spray & pray genning

Anonymous

Anonymous Thu 18 May 2023 21:10:31 No.7382693 Report

Quoted By: >>7382696

>>7382687
Higher batches are actually more efficient per image than lower ones.
>>7382691
I was actually talking about the "constant for dadapt" guy. Dunno where he pulled that one from.

Anonymous

Anonymous Thu 18 May 2023 21:10:37 No.7382694 Report

Quoted By: >>7382700

>>7382690
>no civit
>no huggingface
link your own drive, doubt anything else makes the cut

Anonymous

Anonymous Thu 18 May 2023 21:11:37 No.7382696 Report

Quoted By: >>7382700

>>7382693
constant for dadapt is on said guide and a thing people have been saying for months.

Anonymous

Anonymous Thu 18 May 2023 21:12:35 No.7382698 Report

Quoted By: >>7382702

>>7382671
Yeah, not like the paper has examples of D-Adaptation being used with various schedules, including ones you can't even specify in the script, to reach optimal convergence comparable to hand-tuned LRs.
It literally flat out incorrect to say "D-Adaptation needs constant to work".

Anonymous

Anonymous Thu 18 May 2023 21:13:05 No.7382699 Report

Quoted By: >>7382703

>>7382690
>Make a better template and it will get accepted and used just like this one has.
we had one before someone decided its "outdated" and forced the current garbage. Damn I wonder who could have done that...

Anonymous

Anonymous Thu 18 May 2023 21:13:18 No.7382700 Report

Quoted By: >>7382705 >>7382713 >>7382728

>>7382694
Linking Huggingface is about as good as linking Google. Don't get your point.
Civit is trash enough that not even the last OP filled with shit linked it, so take it up to them.
>>7382696
Then they're shit wrong, because I baked 30 loras on dadapt cosine and they all work well, lmao.

Anonymous

Anonymous Thu 18 May 2023 21:15:10 No.7382702 Report

Quoted By:

>>7382698
depends on the definition of "work" - I assume he meant not being absolute shit, cause sure it'll "work" on anything else but thats not exactly enough.
Im more interested in the arguments he mentions - are they being applied by default on easyscripts or they aren't and that could be why people had shit results with it?

Anonymous

Anonymous Thu 18 May 2023 21:15:31 No.7382703 Report

Quoted By: >>7382713

>>7382699
>and forced
Shit, didn't know I had power over other people baking. I'll make you suck my dick tomorrow.
All this comes down to is you being a lil bitch and not having the balls to bake using the old template, so the previous point still stands.

Anonymous

View Same Google ImgOps iqdb SauceNAO 23689-2809700045-desert sorceres (...).png, 1MiB, 960x960

Anonymous Thu 18 May 2023 21:16:05 No.7382704 Report

Quoted By: >>7382706 >>7382707 >>7382711 >>7382716

Post some fucking gens

Anonymous

Anonymous Thu 18 May 2023 21:16:12 No.7382705 Report

Quoted By: >>7382728 >>7382781

>>7382700
>because I baked 30 loras on dadapt cosine
link?

Anonymous

Anonymous Thu 18 May 2023 21:16:19 No.7382706 Report

Quoted By: >>7382723

>>7382704
post lora

Anonymous

View Same Google ImgOps iqdb SauceNAO 00013-3403560534.png, 2MiB, 1024x1536

Anonymous Thu 18 May 2023 21:17:31 No.7382707 Report

Quoted By:

>>7382704
No. We must fight.

Anonymous

View Same Google ImgOps iqdb SauceNAO catbox_6j018r.png, 2MiB, 1024x1024

Anonymous Thu 18 May 2023 21:18:03 No.7382711 Report

Quoted By: >>7382714 >>7382724

>>7382704

Anonymous

Anonymous Thu 18 May 2023 21:18:51 No.7382712 Report

Quoted By: >>7382726

Expanded some OP links.
https://pastebin.com/0atr0fEc

Anonymous

Anonymous Thu 18 May 2023 21:18:58 No.7382713 Report

Quoted By: >>7382715

>>7382703
anyone who simply bakes first has "power" you imbecile. The only way to avoid this is to start shit flinging with 2 threads which causes mods to come in and that doesn't end well for anyone involved nor the thread.

>>7382700
>Linking Huggingface is about as good as linking Google
so is linking anything at all. are you ok?

Anonymous

View Same Google ImgOps iqdb SauceNAO 06118-1720954395-nsfw, masterpie (...).png, 1007KiB, 768x1152

Anonymous Thu 18 May 2023 21:19:21 No.7382714 Report

Quoted By:

>>7382711
very nice, more like this?

Anonymous

Anonymous Thu 18 May 2023 21:20:05 No.7382715 Report

Quoted By: >>7382720

>>7382713
>anyone who simply bakes first has "power" you imbecile
Then do it first, imbecile.
I didn't even bake the last two threads and the last one I did before that was after two hours of nobody doing it.
The audacity.

Anonymous

View Same Google ImgOps iqdb SauceNAO 1652988915595644.png, 990KiB, 704x896

Anonymous Thu 18 May 2023 21:20:48 No.7382716 Report

Quoted By:

>>7382704

Anonymous

View Same Google ImgOps iqdb SauceNAO g_1684345887856308.jpg, 123KiB, 426x600

Anonymous Thu 18 May 2023 21:21:01 No.7382717 Report

Quoted By:

>>7382686
>Clueless faggot can't provide reasons for his argument on his own

Anonymous

Anonymous Thu 18 May 2023 21:21:58 No.7382720 Report

Quoted By: >>7382721

>>7382715
>then do it first
yes lets race with OPs, great idea.

Anonymous

Anonymous Thu 18 May 2023 21:22:56 No.7382721 Report

Quoted By:

>>7382720
>2 hours downtime
>race
can you kill yourself already natard

Anonymous

Anonymous Thu 18 May 2023 21:25:01 No.7382723 Report

Quoted By:

>>7382706
Should be this one I think https://files.catbox.moe/fmjtlg.safetensors
Be advised that some models really don't like trying to gen characters without visible facial features so you might get varied results based on which checkpoints you're using.

Anonymous

Anonymous Thu 18 May 2023 21:25:26 No.7382724 Report

Quoted By:

>>7382711
Imagine being so low IQ you can't make out this killjoy

Anonymous

View Same Google ImgOps iqdb SauceNAO 01962-1950320470.png, 2MiB, 1536x1024

Anonymous Thu 18 May 2023 21:26:06 No.7382725 Report

Quoted By: >>7383691

>>7382669
>>>/aco/asdg

Anonymous

Anonymous Thu 18 May 2023 21:28:18 No.7382726 Report

Quoted By: >>7382727 >>7382734 >>7382742 >>7382780 >>7383281 >>7383634 >>7386032

>>7382712
>still missing absolutely basic links
why cant you do shit properly.

Would be nice if someone bothered with a small llama install rentry or if there is one already to add alongside web, otherwise here

X Edition

Old: XXX

>WEBUI SETUP
Nvidia: https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Install-and-Run-on-NVidia-GPUs
AMD: https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Install-and-Run-on-AMD-GPUs
Cloud: https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Online-Services
Optimizations: https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Optimizations
Extensions: https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Extensions
Custom Scripts: https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Custom-Scripts

>RESOURCES
Wiki: https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki
LoRAs/models:
https://civitai.com/ | https://huggingface.co/ | https://gitgud.io/gayshit/makesomefuckingporn
Training:
https://github.com/derrian-distro/LoRA_Easy_Training_Scripts | https://rentry.org/59xed3
Tags: https://danbooru.donmai.us/wiki_pages/tag_groups
ControlNet: https://rentry.org/dummycontrolnet
LamaCleaner: https://huggingface.co/spaces/Sanster/Lama-Cleaner-lama
Animation: https://rentry.org/AnimAnon
Wildcards: https://rentry.org/NAIwildcards
4chanX Catbox userscript: https://gist.github.com/catboxanon/ca46eb79ce55e3216aecab49d5c7a3fb
OP Template/Logos: https://pastebin.com/0atr0fEc | https://files.catbox.moe/xk99gy.png | https://files.catbox.moe/35bdt5.gif

Anonymous

Anonymous Thu 18 May 2023 21:29:38 No.7382727 Report

Quoted By:

>>7382726
oops, just fix the newlines when you actually copy it

Anonymous

Anonymous Thu 18 May 2023 21:30:44 No.7382728 Report

Quoted By: >>7382738

>>7382700
>>7382705
>baked 30 loras
>cant link a single one
lol

Anonymous

View Same Google ImgOps iqdb SauceNAO 00228-2402421240.png, 2MiB, 1024x1536

Anonymous Thu 18 May 2023 21:31:38 No.7382730 Report

Quoted By: >>7382732

Anonymous

Anonymous Thu 18 May 2023 21:32:29 No.7382732 Report

Quoted By: >>7382739

>>7382730
catbox my shit up

Anonymous

Anonymous Thu 18 May 2023 21:33:06 No.7382734 Report

Quoted By: >>7382742

>>7382726
See?
It pays off to actually do shit instead of shitting your pants.
I'll add it.

Anonymous

Anonymous Thu 18 May 2023 21:34:45 No.7382738 Report

Quoted By: >>7382745 >>7382749 >>7382759

>>7382728
>know you'll immediately say they're shit without actually using them
>not going to bother
they're on gayshit if you want to test every one ;)

Anonymous

Anonymous Thu 18 May 2023 21:37:08 No.7382739 Report

Quoted By:

>>7382732
Sure. It lost some nice details when I passed it through img2img but I was feeling lazy with this one. I inpainted the face and the entire abdomen from the stomach to the pussy.
https://files.catbox.moe/w21rg5.png

Anonymous

Anonymous Thu 18 May 2023 21:40:39 No.7382742 Report

Quoted By: >>7382748 >>7383634

>>7382726
>>7382734
Rearranged the links just to put our lora resource first.
And linked Lama to their actual site.
Still wouldn't link that guide since it's bloated as shit and has a bunch of bad info, but fuck it, it's the best we have and will probably get since nobody wants to put up.

Anonymous

Anonymous Thu 18 May 2023 21:40:59 No.7382745 Report

Quoted By: >>7382747 >>7382749

>>7382738
Nice larp.
I baked 300 with constant and they're all on gayshit and they all work well too. I also earn $200 form patreon and pixiv corpos tongue my anus.
>:)
of course you're a /g/ tourist

Anonymous

View Same Google ImgOps iqdb SauceNAO 9996490989765785-4057517066-kill (...).png, 2MiB, 1600x2000

Anonymous Thu 18 May 2023 21:41:59 No.7382746 Report

Quoted By: >>7383211

Anonymous

Anonymous Thu 18 May 2023 21:43:09 No.7382747 Report

Quoted By: >>7382751 >>7382759

>>7382745
All good comes to those who are not assblasted.
I'll give you a teeny tiny tease. https://files.catbox.moe/2puud1.png

Anonymous

Anonymous Thu 18 May 2023 21:43:33 No.7382748 Report

Quoted By: >>7382754

>>7382742
>their actual site
does it allow usage in-browser? If you mean github for local install then just link it alongside.

Anonymous

Anonymous Thu 18 May 2023 21:43:41 No.7382749 Report

Quoted By:

>>7382738
>>7382745
lol
I'll just train two LoRAs and do an XYZ, fucking schizos
adamW 1e-4 chads keep winning though

Anonymous

Anonymous Thu 18 May 2023 21:44:34 No.7382751 Report

Quoted By: >>7382754

>>7382747
fuck off back to /g/

Anonymous

View Same Google ImgOps iqdb SauceNAO catbox_1zy633.png, 1MiB, 1024x1024

Anonymous Thu 18 May 2023 21:45:32 No.7382753 Report

Quoted By: >>7382822 >>7383717 >>7384567

Liking all the elf content bros

Anonymous

Anonymous Thu 18 May 2023 21:45:50 No.7382754 Report

Quoted By:

>>7382748
Oh yeah, IG I could put both.
>>7382751
Shit, sorry. Here.
Just so you don't accuse me of using constant.
But wait, I could've inspect elemented it...
https://files.catbox.moe/fettkw.png

Anonymous

View Same Google ImgOps iqdb SauceNAO 00465-2987950408.png, 636KiB, 576x768

Anonymous Thu 18 May 2023 21:45:58 No.7382755 Report

Quoted By:

>>7381766
yea, it's OmarDogan_08v2

Anonymous

Anonymous Thu 18 May 2023 21:48:40 No.7382758 Report

Quoted By:

Good model for having a faceless/out of frame male fucking a girl?

Anonymous

Anonymous Thu 18 May 2023 21:49:31 No.7382759 Report

Quoted By: >>7382760

and we thought the cstaber guy was bad, I'd rather go back to that since he was at least harmless

>>7382738
>>7382747
go back to where you came from please

Anonymous

Anonymous Thu 18 May 2023 21:50:10 No.7382760 Report

Quoted By: >>7383718

>>7382759
>go back to where you came from please
If you pay me 5$, I'll reveal another parameter.
I know, I'm such a tease.

Anonymous

Anonymous Thu 18 May 2023 21:51:07 No.7382761 Report

Quoted By:

dont feed the spammer

Anonymous

Anonymous Thu 18 May 2023 21:56:52 No.7382765 Report

Quoted By: >>7382782

>>7382418
catbox?

Anonymous

View Same Google ImgOps iqdb SauceNAO elf.png, 1MiB, 1024x1024

Anonymous Thu 18 May 2023 21:57:25 No.7382766 Report

Quoted By:

Anonymous

View Same Google ImgOps iqdb SauceNAO 00023-2067735374-1girl, purple e (...).png, 2MiB, 1280x1792

Anonymous Thu 18 May 2023 21:58:59 No.7382767 Report

Quoted By:

Now that we've got that out of our systems, let's get on with it.

Anonymous

Anonymous Thu 18 May 2023 21:59:36 No.7382768 Report

Quoted By: >>7382770 >>7382772

>>7382540
Why does the /e/ thread contain more information then the hdg's op thread?

Anonymous

View Same Google ImgOps iqdb SauceNAO 00681-2699641939.png, 1MiB, 992x992

Anonymous Thu 18 May 2023 22:00:33 No.7382769 Report

Quoted By:

Anonymous

Anonymous Thu 18 May 2023 22:00:34 No.7382770 Report

Quoted By:

>>7382768
The fuck's an /e/

Anonymous

Anonymous Thu 18 May 2023 22:01:15 No.7382772 Report

Quoted By: >>7382775

>>7382768
cause this thread got vandalized. If the autist spammer doesnt ruin it we should have a proper OP again next thread

Anonymous

View Same Google ImgOps iqdb SauceNAO 00001-477438409.png, 2MiB, 1024x1536

Anonymous Thu 18 May 2023 22:01:51 No.7382773 Report

Quoted By:

Anonymous

Anonymous Thu 18 May 2023 22:02:00 No.7382775 Report

Quoted By: >>7382779

>>7382772
Which one's the proper?

Anonymous

View Same Google ImgOps iqdb SauceNAO 44336-770654025-best quality, fa (...).png, 1MiB, 1024x1280

Anonymous Thu 18 May 2023 22:02:12 No.7382776 Report

Quoted By:

Anonymous

Anonymous Thu 18 May 2023 22:04:43 No.7382779 Report

Quoted By: >>7382780

>>7382775
scroll up

Anonymous

Anonymous Thu 18 May 2023 22:05:59 No.7382780 Report

Quoted By:

>>7382779
>>7382726 this?

Anonymous

Anonymous Thu 18 May 2023 22:06:04 No.7382781 Report

Quoted By: >>7382784 >>7382787

>>7382705
hey its this retard again who challenged my to link my loras then never said another word after i did lmao

Anonymous

Anonymous Thu 18 May 2023 22:07:02 No.7382782 Report

Quoted By:

>>7382765
https://files.catbox.moe/hkdiec.png

Anonymous

Anonymous Thu 18 May 2023 22:08:31 No.7382784 Report

Quoted By:

>>7382781
whats there to say? the point is to prove that you actually baked shit and not just talking out of your ass, what are you on about

Anonymous

Anonymous Thu 18 May 2023 22:09:41 No.7382787 Report

Quoted By:

>>7382781
awfully smug for someone who admitted his settings were shit when pointed out in that same thread after someone actually used them

Anonymous

View Same Google ImgOps iqdb SauceNAO 00097-1200870648-nsfw, scan, (sk (...).png, 2MiB, 1024x1536

Anonymous Thu 18 May 2023 22:12:42 No.7382789 Report

Quoted By:

Anonymous

View Same Google ImgOps iqdb SauceNAO file.png, 87KiB, 661x716

Anonymous Thu 18 May 2023 22:14:13 No.7382791 Report

Quoted By: >>7382799 >>7382816 >>7382886

Anonymous

View Same Google ImgOps iqdb SauceNAO 01148-2208735274-ghibli_style, 1 (...).png, 1MiB, 1024x1536

Anonymous Thu 18 May 2023 22:14:32 No.7382792 Report

Quoted By: >>7382793 >>7382794 >>7382795 >>7382796 >>7382798 >>7382856

new to proompting

Anonymous

Anonymous Thu 18 May 2023 22:15:24 No.7382793 Report

Quoted By:

>>7382792
>already better than most of /hdg/
based

Anonymous

Anonymous Thu 18 May 2023 22:15:41 No.7382794 Report

Quoted By:

>>7382792
nice

Anonymous

Anonymous Thu 18 May 2023 22:16:06 No.7382795 Report

Quoted By:

>>7382792
It's fucking nice.

Anonymous

Anonymous Thu 18 May 2023 22:16:15 No.7382796 Report

Quoted By: >>7382803

>>7382792
she has some cola on her boobers

Anonymous

Anonymous Thu 18 May 2023 22:17:12 No.7382798 Report

Quoted By:

>>7382792
boob

Anonymous

Anonymous Thu 18 May 2023 22:17:22 No.7382799 Report

Quoted By: >>7382820

>>7382791
>think about moving somewhere else because it's so terrible here now
>the best alternative is filled with bears sucking cocks
hmm...

Anonymous

View Same Google ImgOps iqdb SauceNAO gorgeouswifey.png, 2MiB, 2816x2248

Anonymous Thu 18 May 2023 22:18:15 No.7382800 Report

Quoted By: >>7382802 >>7382805

btw ty for the help before anons, still no lora for her but I'm figuring out how to gen and inpaint my tulpa wife more consistently and I am happy

Anonymous

View Same Google ImgOps iqdb SauceNAO tiddies.png, 2MiB, 2944x2355

Anonymous Thu 18 May 2023 22:19:29 No.7382802 Report

Quoted By:

>>7382800

Anonymous

Anonymous Thu 18 May 2023 22:19:37 No.7382803 Report

Quoted By: >>7382804 >>7382808

>>7382796
yeah no idea how to do impainting. maybe i should just photoshop it

Anonymous

Anonymous Thu 18 May 2023 22:20:11 No.7382804 Report

Quoted By:

>>7382803
given the flat shading that looks pretty easy

Anonymous

Anonymous Thu 18 May 2023 22:20:29 No.7382805 Report

Quoted By: >>7382810 >>7382847

>>7382800
>still no lora for her
Did you try making one yet or are you still collecting images? I remember you from one of the earlier threads
Kinda curious how you did the hair streaks, coincidentally I was thinking about them earlier but I didn't get around to see if the AI recognises it as a tag

Anonymous

Anonymous Thu 18 May 2023 22:21:11 No.7382808 Report

Quoted By:

>>7382803
>put it in lama
>3s later
>img2img at low denoise to unify and preserve metadata

Anonymous

View Same Google ImgOps iqdb SauceNAO pair_of_boobas.png, 1MiB, 992x992

Anonymous Thu 18 May 2023 22:21:18 No.7382809 Report

Quoted By: >>7383182

Anonymous

Anonymous Thu 18 May 2023 22:21:53 No.7382810 Report

Quoted By: >>7382812

>>7382805
I'm just inpainting them. I realized what I was fucking up with inpainting that wasn't working for me before so now I can just take these, MSpaint in some color streaks, and then inpaint it back in. One of the easiest things to inpaint quickly, actually

Anonymous

Anonymous Thu 18 May 2023 22:23:30 No.7382812 Report

Quoted By:

>>7382810
I do the same, it's a lot easier to nudge the inpainting tool towards whatever you had in mind if you give it something to try to modify

Anonymous

View Same Google ImgOps iqdb SauceNAO file.png, 7KiB, 223x213

Anonymous Thu 18 May 2023 22:24:29 No.7382816 Report

Quoted By: >>7382821 >>7382828

>>7382791
isnt this supposed to be filled with insane shit? whats wrong with the xyz or adam?

Anonymous

Anonymous Thu 18 May 2023 22:25:10 No.7382818 Report

Quoted By: >>7382844

>>7379902
Catbox?

Anonymous

Anonymous Thu 18 May 2023 22:25:30 No.7382820 Report

Quoted By: >>7382824 >>7382850 >>7382889

>>7382799
The worst part is im starting to consider the furryshit place more and more each day cause it just keeps getting worse here.

Anonymous

Anonymous Thu 18 May 2023 22:25:37 No.7382821 Report

Quoted By: >>7382827

>>7382816
It's referring to the gatekeeper who also tests his loras on 0.1 stength.

Anonymous

Anonymous Thu 18 May 2023 22:25:47 No.7382822 Report

Quoted By:

>>7382753
For some reason I just felt like it today. Maybe I'm too used to seeing big titty elves nowadays.

Anonymous

Anonymous Thu 18 May 2023 22:26:38 No.7382824 Report

Quoted By:

>>7382820
Honestly. They put in some work.
Maybe we should leave them alone to stop contamination by jaded hdgfags.

Anonymous

Anonymous Thu 18 May 2023 22:27:00 No.7382826 Report

Quoted By:

Start talking about money.

Anonymous

Anonymous Thu 18 May 2023 22:27:09 No.7382827 Report

Quoted By:

>>7382821
that doesnt even sound related

Anonymous

Anonymous Thu 18 May 2023 22:27:10 No.7382828 Report

Quoted By:

>>7382816
Most of those were just the first thing that came to mind t b h, it's such a recurring discussion that I feel like I read the same thing all over again multiple times per week

Anonymous

View Same Google ImgOps iqdb SauceNAO 00002-421801101.png, 2MiB, 1024x1536

Anonymous Thu 18 May 2023 22:27:32 No.7382830 Report

Quoted By:

Anonymous

View Same Google ImgOps iqdb SauceNAO catbox_4wwe88.png, 1MiB, 1024x1024

Anonymous Thu 18 May 2023 22:28:04 No.7382831 Report

Quoted By: >>7382834 >>7382854 >>7382886 >>7383111

Its honestly not that bad here

Anonymous

Anonymous Thu 18 May 2023 22:29:58 No.7382834 Report

Quoted By:

>>7382831
:D

Anonymous

Anonymous Thu 18 May 2023 22:30:54 No.7382835 Report

Quoted By:

adamw virgins
dadapt goobers
lion archons
adafactor patricians

Anonymous

View Same Google ImgOps iqdb SauceNAO 01153-2394050289-1girl, jar jar (...).png, 1MiB, 1024x1536

Anonymous Thu 18 May 2023 22:33:43 No.7382837 Report

Quoted By: >>7382838

Anonymous

Anonymous Thu 18 May 2023 22:33:54 No.7382838 Report

Quoted By:

>>7382837
BASED

Anonymous

View Same Google ImgOps iqdb SauceNAO EP1-4118.jpg, 315KiB, 1920x1080

Anonymous Thu 18 May 2023 22:34:04 No.7382839 Report

Quoted By: >>7382841 >>7382886

Do you want to include or exclude group shots like these when training a style LoRA?

Anonymous

Anonymous Thu 18 May 2023 22:35:47 No.7382841 Report

Quoted By: >>7382845

>>7382839
my favorite shitposter.

Anonymous

View Same Google ImgOps iqdb SauceNAO 9996490989765781-1007540802-kill (...).png, 2MiB, 1600x2000

Anonymous Thu 18 May 2023 22:43:08 No.7382844 Report

Quoted By:

>>7382818
here you go. i used controlnet but i dont remember original picture.
edited.
https://files.catbox.moe/tjip8k.png
original
https://files.catbox.moe/mv3xif.png

Anonymous

Anonymous Thu 18 May 2023 22:44:05 No.7382845 Report

Quoted By: >>7382855

>>7382841
genuine question

Anonymous

View Same Google ImgOps iqdb SauceNAO riding.png, 3MiB, 2048x2880

Anonymous Thu 18 May 2023 22:44:29 No.7382847 Report

Quoted By:

>>7382805
And yes, still making images

Anonymous

Anonymous Thu 18 May 2023 22:45:41 No.7382850 Report

Quoted By:

>>7382820
It's tempting because they're comfy, but I don't want to ruin their fun.

Anonymous

View Same Google ImgOps iqdb SauceNAO 00238-2576876893.png, 2MiB, 1536x1024

Anonymous Thu 18 May 2023 22:45:50 No.7382851 Report

Quoted By:

Anonymous

Anonymous Thu 18 May 2023 22:46:52 No.7382854 Report

Quoted By: >>7383014

>>7382831
kids off school for the spring its a sitewide problem

Anonymous

Anonymous Thu 18 May 2023 22:47:37 No.7382855 Report

Quoted By: >>7382858 >>7382860

>>7382845
the previous few times it was too hilarious to bother answering, but sure, seems legit
the answer is that no one knows. I mean, people tend to suggest only leaving the 1 character but its not like anyone really tested the exact effect it has. It also depends on how big the dataset is and what portion of it will be group images. And no there is no solid answers as to what the ratio should be for this either.

Anonymous

View Same Google ImgOps iqdb SauceNAO 01146-1292332875-4k, intricate d (...).png, 2MiB, 1024x1536

Anonymous Thu 18 May 2023 22:49:42 No.7382856 Report

Quoted By: >>7383015 >>7383046

>>7382792
something else i tried

Anonymous

Anonymous Thu 18 May 2023 22:51:49 No.7382858 Report

Quoted By: >>7382863

>>7382855
main reason for wanting to include them is that these are the ones that most often include (nearly) full-body shots of the characters and I don't want to crop them for each one

Anonymous

Anonymous Thu 18 May 2023 22:52:03 No.7382860 Report

Quoted By: >>7382863

>>7382855
he said for a style lora though and i cant see why it would matter for a style

Anonymous

Anonymous Thu 18 May 2023 22:54:03 No.7382863 Report

Quoted By:

>>7382860
oh. yeah im tired. >>7382858 if its for style then sure you should be including these. Random untagged objects and busy backgrounds seem to be more of an issue from my experience

Anonymous

Anonymous Thu 18 May 2023 22:54:47 No.7382864 Report

Quoted By: >>7382868 >>7382935 >>7383273

>hdg: zero news and innovation; constant arguing as result; nude version of sdg; 1girl, smile
>sdg: sfw lmao
>vtai: sfw lmao
>csdg: 3dpd fugly
>ddg: exclusively futashit and other ugliness
>b: no comment
>trash sdg: pretty nice but exclusively furry
>e sdg: 7 day old thread
it's owari

Anonymous

Anonymous Thu 18 May 2023 22:57:33 No.7382865 Report

Quoted By: >>7382866 >>7382868

As a returning proompter, which is the best?
>Automatic1111 Webui
>Vladmandic
>Webui-ux

Anonymous

Anonymous Thu 18 May 2023 22:58:32 No.7382866 Report

Quoted By:

>>7382865
auto

Anonymous

Anonymous Thu 18 May 2023 23:01:01 No.7382868 Report

Quoted By: >>7382869 >>7382872

>>7382865
#1

>>7382864
dont you ever get tired of dooming holy shit its so fucking tedious, i cant imagine making posts like this twice a day for months

Anonymous

Anonymous Thu 18 May 2023 23:02:08 No.7382869 Report

Quoted By: >>7382872

>>7382868
>spent months on dooming already
it's so over...

Anonymous

Anonymous Thu 18 May 2023 23:04:42 No.7382872 Report

Quoted By:

>>7382868
>pixiv banning requests
>fanbox banning ai
if there was a time to doom the time is now.

>>7382869
>hes dooming for months and its getting worse despite that
truly over it has become...

Anonymous

Anonymous Thu 18 May 2023 23:07:28 No.7382877 Report

Quoted By:

>decensor inpaint doesn't work using 'whole picture' but only on one specific picture
This shit will give an aunerysm some day.

Anonymous

View Same Google ImgOps iqdb SauceNAO 00049-1607450346-1girl, short ha (...).png, 2MiB, 1288x1792

Anonymous Thu 18 May 2023 23:08:48 No.7382881 Report

Quoted By: >>7382903

Anonymous

View Same Google ImgOps iqdb SauceNAO 1673603188801962.png, 2MiB, 1024x1280

Anonymous Thu 18 May 2023 23:11:34 No.7382886 Report

Quoted By:

>>7382791
kek
>>7382831
Whoa, anon. That's not very doomposting of you
>>7382839
From my experience it's generally fine as long as you don't have too many images that have characters in unreadable/unclear positions that could mess up anatomy. Only the hand position of the two center characters are a bit rough, but if it's tagged correctly, it should be ok. Alternatively you could try to make some rough crops where it makes sense.

Anonymous

View Same Google ImgOps iqdb SauceNAO 06740-289850530-nsfw, masterpiec (...).png, 996KiB, 768x1152

Anonymous Thu 18 May 2023 23:11:40 No.7382887 Report

Quoted By:

Anonymous

View Same Google ImgOps iqdb SauceNAO 00076.jpg, 1MiB, 2003x3000

Anonymous Thu 18 May 2023 23:12:14 No.7382888 Report

Quoted By: >>7382903 >>7382905

AI is pretty neat

Anonymous

Anonymous Thu 18 May 2023 23:12:28 No.7382889 Report

Quoted By:

>>7382820
/trash/ isn't much better honestly
the most comfy one is /vt/, but you gotta proompt those

Anonymous

View Same Google ImgOps iqdb SauceNAO 00305-2393675149.jpg, 4MiB, 3828x1488

Anonymous Thu 18 May 2023 23:13:57 No.7382892 Report

Quoted By: >>7382898 >>7382902

Anonymous

Anonymous Thu 18 May 2023 23:15:15 No.7382893 Report

Quoted By: >>7382895 >>7382899 >>7382900 >>7382902 >>7382903

Artificial Academy 2 General /aa2g/ #1279
Who Framed /aa2g/? Edition

Welcome, this general is for the discussion of ILLUSION's Artificial Academy 2.

COPY ERROR MESSAGES WITH CTRL+C, PASTE THEM WITH CTRL+V INTO GOOGLE TRANSLATE. JUST CLICK THE WINDOW AND PRESS CTRL + C, IT WORKS.

>Downloads:
/aa2g/ Pre-Installed Game, AA2Mini: https://tsukiyo.me/AAA/AA2MiniPPX.xml
AAUnlimited updates: https://github.com/aa2g/AA2Unlimited/releases
Anon's Modded Pre-Install: https://pastebin.com/42JS3q6E

>Information:
AA2Mini Install Guide:
https://docs.google.com/document/d/e/2PACX-1vS8Ap6CrmSNXRsKG9jsIMqHYuHM3Cfs5qE5nX6iIgfzLlcWnmiwzmOrp27ytEMX03lFNRR7U5UXJalA/pub
General FAQ:
https://web.archive.org/web/20200216045726/https://pastebin.com/bhrA6iGx
AAU Guide and Resources (Modules, Tans, Props, Poses, and More):
https://docs.google.com/spreadsheets/d/17qb1X0oOdMKU4OIDp8AfFdLtl5y_4jeOOQfPQ2F-PKQ/edit#gid=0

>Character Cards [Database], now with a list of every NonOC in the megas:
https://docs.google.com/spreadsheets/d/1niC6g-Xd2a2yaY98NBFdAXnURi4ly2-lKty69rkQbJ0/edit#gid=2085826690
https://db.bepis.moe/aa2/

>Mods & More:
Mods for AAU/AA2Mini (ppx format, the mediafire has everything):
https://www.mediafire.com/folder/vwrmdohus4vhh/Mods
/aa2g/ Modding Reference Guide (Slot lists for Hair/Clothes/Faces, List Guides, and More):
https://docs.google.com/spreadsheets/d/1gwmoVpKuSuF0PtEPLEB17eK_dexPaKU106ShZEpBLhg/edit#gid=1751233129
Booru: https://aau.booru.org

>HELP! I have a Nvidia card and my game crashes on startup!
Try the dgVoodoo option in the new win10fix settings.
Alternative: Update your AAU and see if it happens again. If so, disable win10fix, enable wined3d and software vertex processing.
>HELP! Required Windows 11 update broke things!
winkey+R -> ms-settings:developers -> Terminal=Windows Console Host

Previous Thread:
>>428858839

Anonymous

Anonymous Thu 18 May 2023 23:15:59 No.7382895 Report

Quoted By: >>7382900

>>7382893
a-anon?

Anonymous

Anonymous Thu 18 May 2023 23:16:04 No.7382896 Report

Quoted By: >>7382901

>batch size: 2
>oom
>2 batches with batch size 1
>ok
Hackerman

Anonymous

View Same Google ImgOps iqdb SauceNAO 00274-2730313661.png, 2MiB, 1536x1024

Anonymous Thu 18 May 2023 23:16:25 No.7382898 Report

Quoted By:

>>7382892
kek

Anonymous

Anonymous Thu 18 May 2023 23:17:05 No.7382899 Report

Quoted By:

>>7382893
based, please stay

Anonymous

Anonymous Thu 18 May 2023 23:17:36 No.7382900 Report

Quoted By: >>7382904

>>7382893
>>7382895
what did he mean by this?

Anonymous

View Same Google ImgOps iqdb SauceNAO 06871-1180475620-masterpiece, be (...).png, 898KiB, 768x1152

Anonymous Thu 18 May 2023 23:17:53 No.7382901 Report

Quoted By:

>>7382896
it just works

Anonymous

Anonymous Thu 18 May 2023 23:17:55 No.7382902 Report

Quoted By: >>7382985

>>7382892
I don't know what drug you're on, but I want some.
>>7382893
go home /aa2g/, you're drunk

Anonymous

View Same Google ImgOps iqdb SauceNAO 00058-3741541165-1girl, mature f (...).png, 2MiB, 1280x1792

Anonymous Thu 18 May 2023 23:17:58 No.7382903 Report

Quoted By:

>>7382881
another
>>7382888
What is this, a picture for boomers?!
>>7382893
based retard

Anonymous

Anonymous Thu 18 May 2023 23:18:17 No.7382904 Report

Quoted By: >>7382910

>>7382900
he does this to a few boards every few day
probably trying to save his dead thread

Anonymous

Anonymous Thu 18 May 2023 23:19:09 No.7382905 Report

Quoted By:

>>7382888
Based pixelchad

Anonymous

Anonymous Thu 18 May 2023 23:20:49 No.7382910 Report

Quoted By:

>>7382904
oh, I think I saw it somewhere before now that you mention it

Anonymous

View Same Google ImgOps iqdb SauceNAO 00023-3059268429.png, 2MiB, 1600x1024

Anonymous Thu 18 May 2023 23:30:08 No.7382915 Report

Quoted By: >>7382916 >>7382921

10 inches

Anonymous

Anonymous Thu 18 May 2023 23:31:56 No.7382916 Report

Quoted By:

>>7382915
This is what guys mean when they say they're hung like a horse.

Anonymous

Anonymous Thu 18 May 2023 23:34:33 No.7382917 Report

Quoted By:

>concept lora completely obliterates image
sometimes i wonder how these people trained those

Anonymous

Anonymous Thu 18 May 2023 23:36:54 No.7382921 Report

Quoted By: >>7382923

>>7382915
what a monster

Anonymous

Anonymous Thu 18 May 2023 23:38:26 No.7382922 Report

Quoted By: >>7382924

should i be using xformers?

Anonymous

View Same Google ImgOps iqdb SauceNAO 00001-1048340161.png, 510KiB, 800x512

Anonymous Thu 18 May 2023 23:39:20 No.7382923 Report

Quoted By:

>>7382921
looked fine before the upscaling, won't complain, i take that saliva coated horse dick for my image haha

Anonymous

Anonymous Thu 18 May 2023 23:39:20 No.7382924 Report

Quoted By: >>7382941

>>7382922
Nah, the 35% speed increase is not desired.

Anonymous

Anonymous Thu 18 May 2023 23:40:22 No.7382927 Report

Quoted By:

>blowie close up
>mentioning clothes or nudity fucks up the close up
>not mentioning clothes makes monstrosities
owari

Anonymous

Anonymous Thu 18 May 2023 23:54:16 No.7382935 Report

Quoted By: >>7382937 >>7382944

>>7382864
Is it over, AIbros? Are we hitting stagnation?

Anonymous

Anonymous Thu 18 May 2023 23:55:00 No.7382937 Report

Quoted By:

>>7382935
we're hitting deez

Anonymous

Anonymous Thu 18 May 2023 23:55:11 No.7382938 Report

Quoted By: >>7382939

>1girl, smile

Whats wrong with this?

Anonymous

Anonymous Thu 18 May 2023 23:55:43 No.7382939 Report

Quoted By:

>>7382938
nothing

Anonymous

Anonymous Thu 18 May 2023 23:56:34 No.7382941 Report

Quoted By: >>7382942

>>7382924
Not that anon but I updated my graphic card drives and now xformers its fucked

Anonymous

Anonymous Thu 18 May 2023 23:58:47 No.7382942 Report

Quoted By:

>>7382941
Meant to say drivers of course

Anonymous

Anonymous Thu 18 May 2023 23:59:40 No.7382944 Report

Quoted By:

>>7382935
>still the same model since november
the last ai winter lasted over 20 years haha...

Anonymous

View Same Google ImgOps iqdb SauceNAO afterschoolselfie.png, 4MiB, 1383x1944

Anonymous Fri 19 May 2023 00:00:07 No.7382946 Report

Quoted By:

Trying out 7th Layer model

Anonymous

Anonymous Fri 19 May 2023 00:11:52 No.7382954 Report

Quoted By: >>7382955

Why don't people making LoRA merged models just release them as LoRAs? They don't even tell the base model they merged with so I can't do a clean extraction either, though it seems to work out most of the time regardless ig

Anonymous

Anonymous Fri 19 May 2023 00:12:53 No.7382955 Report

Quoted By: >>7382965

>>7382954
The only people who do that are from Civit, so here's your answer.

Anonymous

Anonymous Fri 19 May 2023 00:24:10 No.7382965 Report

Quoted By:

>>7382955
The other places are doing fuck all so I kinda don't care tbdesu. I'm curious what fucked up YouTube tutorial they're following now that teaches them to train a LoRA and merge instead of just training the model directly.

Anonymous

View Same Google ImgOps iqdb SauceNAO 00079-2558734884-1girl, mature f (...).png, 2MiB, 1280x1792

Anonymous Fri 19 May 2023 00:40:45 No.7382980 Report

Quoted By:

Anonymous

Anonymous Fri 19 May 2023 00:41:19 No.7382981 Report

Quoted By: >>7382986

>download 4k images from an artist
>after getting rid of duplicates only about 600 images remain
>know this group likely contains images of both his new and old artstyle so I'll probably need to do lots of manual sorting
>will likely also have to take out a few trash pics here and there
This is by far the most tedious part of baking

Anonymous

View Same Google ImgOps iqdb SauceNAO 00026-2338800790.png, 2MiB, 1264x920

Anonymous Fri 19 May 2023 00:42:59 No.7382984 Report

Quoted By:

Anonymous

View Same Google ImgOps iqdb SauceNAO 01100100 01110010 01110101 01100 (...).jpg, 3MiB, 2808x1488

Anonymous Fri 19 May 2023 00:44:33 No.7382985 Report

Quoted By:

>>7382902

Anonymous

Anonymous Fri 19 May 2023 00:44:50 No.7382986 Report

Quoted By: >>7382987

>>7382981
I don't mind the sorting. It does quick of your image editor supports instadelete.

Anonymous

Anonymous Fri 19 May 2023 00:45:51 No.7382987 Report

Quoted By:

>>7382986
goes quick*
if*

Anonymous

View Same Google ImgOps iqdb SauceNAO catbox_esbda4.png, 786KiB, 768x768

Anonymous Fri 19 May 2023 00:48:46 No.7382992 Report

Quoted By:

Fellas, which users have fanboxes/patreons worth checking with lora and other SD resources?

Anonymous

Anonymous Fri 19 May 2023 01:20:55 No.7383009 Report

Quoted By: >>7383048

>>7382639
>nobody would be against putting it in, desu.
i have seen multiple people complain about your retarded op already. it's more like you're the only one that would have a problem putting it in

Anonymous

View Same Google ImgOps iqdb SauceNAO sunday-best.png, 1MiB, 1228x1574

Anonymous Fri 19 May 2023 01:22:12 No.7383010 Report

Quoted By:

wifey likes to go to church with me

Anonymous

View Same Google ImgOps iqdb SauceNAO 00034-1161164610.png, 2MiB, 1920x1080

Anonymous Fri 19 May 2023 01:24:10 No.7383013 Report

Quoted By:

Attempt to make a 1080p wallpaper end up with a fucked up house interior

Anonymous

Anonymous Fri 19 May 2023 01:24:57 No.7383014 Report

Quoted By: >>7383605

>>7382854
>spring
>schizo posting only started in the spring
how new are you

Anonymous

Anonymous Fri 19 May 2023 01:26:04 No.7383015 Report

Quoted By: >>7383038 >>7383188

>>7382856
is that the civitai arcane lora? god you can just see how overfit it is from the hair

Anonymous

View Same Google ImgOps iqdb SauceNAO 06957-4168185252-masterpiece, be (...).png, 856KiB, 768x1152

Anonymous Fri 19 May 2023 01:42:34 No.7383031 Report

Quoted By:

your desert, sir

Anonymous

View Same Google ImgOps iqdb SauceNAO 00116-4165013105.png, 2MiB, 1536x1024

Anonymous Fri 19 May 2023 01:42:52 No.7383032 Report

Quoted By: >>7383035

Trying to use the cyborg LoRA to prompt Cibo from BLAME!.

Anonymous

View Same Google ImgOps iqdb SauceNAO 00129-1131575854.png, 2MiB, 1536x1024

Anonymous Fri 19 May 2023 01:45:09 No.7383035 Report

Quoted By: >>7383045

>>7383032
Trying to get BLAME!-style backgrounds is an interesting challenge. The baseline Stable Diffusion model recognizes Tsutomu Nihei as a token and knows what his style is like.

Anonymous

Anonymous Fri 19 May 2023 01:45:47 No.7383038 Report

Quoted By:

>>7383015
I don't see no desert there...

Anonymous

View Same Google ImgOps iqdb SauceNAO 00282-4251724564.png, 2MiB, 1536x1024

Anonymous Fri 19 May 2023 01:49:10 No.7383045 Report

Quoted By: >>7383049 >>7383202

>>7383035
Eventually, the prompt I settled on using for the background is:
(very wide angle:1.2), (dramatic lighting, <gigantic immense cavernous concrete brutalist megastructure> interior with bottomless pits:1.7) by Tsutomu Nihei

Anonymous

Anonymous Fri 19 May 2023 01:49:20 No.7383046 Report

Quoted By:

>>7382856
yeah it is. But pink hair was in the prompt

Anonymous

Anonymous Fri 19 May 2023 01:50:03 No.7383048 Report

Quoted By:

>>7383009
I do, mhm. It's still in tho.

Anonymous

View Same Google ImgOps iqdb SauceNAO 00322-2134046712.png, 2MiB, 1536x1024

Anonymous Fri 19 May 2023 01:50:07 No.7383049 Report

Quoted By: >>7383202

>>7383045
It's a pity that prompting for sex seems to make the backgrounds less interesting.

Anonymous

Anonymous Fri 19 May 2023 02:03:05 No.7383058 Report

Quoted By: >>7383059

Whats a good alternative to Based64 V3 I tried anything-v4.0 only to get mixed results also tried AOM3A3 with also mixed results

Anonymous

Anonymous Fri 19 May 2023 02:03:44 No.7383059 Report

Quoted By: >>7383060

>>7383058
>alternative
but why

Anonymous

Anonymous Fri 19 May 2023 02:05:04 No.7383060 Report

Quoted By: >>7383077

>>7383059
I mean using the same model over and over again is boring gotta try something new you know

Anonymous

View Same Google ImgOps iqdb SauceNAO 22944-2240577898.jpg, 486KiB, 997x905

Anonymous Fri 19 May 2023 02:09:37 No.7383069 Report

Quoted By: >>7383072 >>7383073 >>7383075

>>>/aco/7278765

Anonymous

View Same Google ImgOps iqdb SauceNAO 22932-1553553989.png, 2MiB, 1408x1408

Anonymous Fri 19 May 2023 02:11:07 No.7383071 Report

Quoted By: >>7383075 >>7383082

random glowing cattails

Anonymous

Anonymous Fri 19 May 2023 02:11:53 No.7383072 Report

Quoted By:

>>7383069
pedo nigger gtfo

Anonymous

Anonymous Fri 19 May 2023 02:12:24 No.7383073 Report

Quoted By:

>>7383069
Hello, Special Agent Johnson!

Anonymous

Anonymous Fri 19 May 2023 02:12:39 No.7383075 Report

Quoted By:

>>7383069
>>7383071
fuck off retard

Anonymous

Anonymous Fri 19 May 2023 02:13:47 No.7383077 Report

Quoted By:

>>7383060
Oh. I use aom2strawberry. It's pretty similar to B64 though.
There's nothing interesting on the mix front.

Anonymous

Anonymous Fri 19 May 2023 02:14:54 No.7383080 Report

Quoted By:

so in the report window we select "This post violates United States law." right?
I think 3dcg falls under this under u.s. law

Anonymous

Anonymous Fri 19 May 2023 02:15:42 No.7383082 Report

Quoted By:

>>7383071
This post literally glows.

lolicon

View Same Google ImgOps iqdb SauceNAO 22928-3446086588.jpg, 707KiB, 1408x1302

lolicon Fri 19 May 2023 02:16:20 No.7383086 Report

Quoted By:

its just me gawd cant you take it easy just enjoy the loli

Anonymous

View Same Google ImgOps iqdb SauceNAO 01291-1898298596-nsfw, uncensore (...).png, 980KiB, 768x1024

Anonymous Fri 19 May 2023 02:18:11 No.7383088 Report

Quoted By:

lolicon

View Same Google ImgOps iqdb SauceNAO 22915-2384534005.jpg, 577KiB, 1408x1014

lolicon Fri 19 May 2023 02:18:51 No.7383089 Report

Quoted By: >>7383094

Anonymous

View Same Google ImgOps iqdb SauceNAO 00001.png, 2MiB, 1536x2048

Anonymous Fri 19 May 2023 02:19:59 No.7383092 Report

Quoted By:

Anonymous

Anonymous Fri 19 May 2023 02:20:25 No.7383094 Report

Quoted By: >>7383095

>>7383089
actually kys

i'll be back when this inbred is gone

lolicon

View Same Google ImgOps iqdb SauceNAO 22913-4029503330.jpg, 940KiB, 1408x1242

lolicon Fri 19 May 2023 02:21:28 No.7383095 Report

Quoted By:

>>7383094
so hostile to loli ;-;

lolicon

View Same Google ImgOps iqdb SauceNAO 22911-3873388276.jpg, 510KiB, 1408x800

lolicon Fri 19 May 2023 02:22:36 No.7383100 Report

Quoted By:

Anonymous

Anonymous Fri 19 May 2023 02:24:03 No.7383104 Report

Quoted By:

>absolute state of this general
lmao
maybe bear fellatio isn't so bad

lolicon

View Same Google ImgOps iqdb SauceNAO 22905-2072185247.jpg, 630KiB, 1408x1080

lolicon Fri 19 May 2023 02:26:11 No.7383109 Report

Quoted By: >>7383111

Anonymous

View Same Google ImgOps iqdb SauceNAO 00113-2774506894 (1).png, 1MiB, 1024x1536

Anonymous Fri 19 May 2023 02:26:19 No.7383110 Report

Quoted By:

takes a bit of prompting to get that outfit, i think maybey i should train a specific prompt from it.

Anonymous

Anonymous Fri 19 May 2023 02:29:07 No.7383111 Report

Quoted By: >>7383117

>>7382831
>not that bad here
>>>7383109

Anonymous

Anonymous Fri 19 May 2023 02:30:12 No.7383114 Report

Quoted By: >>7383135

how do you explain stable diffusion to your parents?

Anonymous

View Same Google ImgOps iqdb SauceNAO catbox_6wbw0j.png, 1MiB, 1024x1024

Anonymous Fri 19 May 2023 02:31:11 No.7383117 Report

Quoted By: >>7383127

>>7383111
1 shitposter is so w/e

lolicon

View Same Google ImgOps iqdb SauceNAO 22890-1363123684.jpg, 516KiB, 1408x852

lolicon Fri 19 May 2023 02:31:12 No.7383118 Report

Quoted By:

ye we chillin

Anonymous

Anonymous Fri 19 May 2023 02:33:10 No.7383126 Report

Quoted By: >>7383163

https://civitai.com/models/65871/lakilolom-style
Did any of you manage to grab this? It's been deleted.

Anonymous

Anonymous Fri 19 May 2023 02:34:56 No.7383127 Report

Quoted By: >>7383132

>>7383117
cp is w/e now
it's over

lolicon

View Same Google ImgOps iqdb SauceNAO 22869-2493750403.jpg, 900KiB, 1408x1408

lolicon Fri 19 May 2023 02:35:13 No.7383129 Report

Quoted By: >>7383130

im am seriously concerned with how hostile people are these days, can you believe want to murder you for sharing cute things?

oh my god this is absurd

Anonymous

Anonymous Fri 19 May 2023 02:36:56 No.7383130 Report

Quoted By: >>7383136

>>7383129
there's no way this thing isn't mentally ill, trolling or both

Anonymous

Anonymous Fri 19 May 2023 02:37:26 No.7383132 Report

Quoted By: >>7383137

>>7383127
What do you want? Its the mods job

Anonymous

Anonymous Fri 19 May 2023 02:38:24 No.7383135 Report

Quoted By:

>>7383114
it generates pictures when you type stuff in

Anonymous

Anonymous Fri 19 May 2023 02:39:25 No.7383136 Report

Quoted By: >>7383139

>>7383130
Do you know how boring it gets at NSA?

Anonymous

Anonymous Fri 19 May 2023 02:39:30 No.7383137 Report

Quoted By:

>>7383132
>What do you want?
we could try doxxing the guy.

Anonymous

Anonymous Fri 19 May 2023 02:40:24 No.7383139 Report

Quoted By:

>>7383136
>What do you want?
i imagine its like the movie burn after reading.

Anonymous

Anonymous Fri 19 May 2023 02:44:17 No.7383146 Report

Quoted By: >>7383147 >>7383149

cp fag is kill
hooray

Anonymous

Anonymous Fri 19 May 2023 02:46:14 No.7383147 Report

Quoted By:

>>7383146
wowie 3 days again
you get more for saying you're 17

lolicon

View Same Google ImgOps iqdb SauceNAO 22983-1016569036.jpg, 866KiB, 1408x1064

lolicon Fri 19 May 2023 02:49:31 No.7383149 Report

Quoted By:

>>7383146
hooray cp is kill
fag

lolicon

View Same Google ImgOps iqdb SauceNAO 23000-1821057573.jpg, 642KiB, 1408x988

lolicon Fri 19 May 2023 02:54:20 No.7383152 Report

Quoted By:

heres the prompt make your own

pos:
little girls,nude,(detailed),spread legs,blue-pink hair,pale white skin,smile,(darkness),glowing eyes,thigh focus,twin tail,bows,(kissing),skinny,mesomorph,choker

neg:
fat belly,fat stomach,fat abdomen,thick,wide hips,large hips,large breasts,large head,shirt,large hands,large fingers,grey,plastic,smirk,blue skin

Steps: 20, Sampler: Euler a, CFG scale: 7, Seed: -1, Size: 1408x1408, Model hash: my secret recipe, Denoising strength: 0.6, First pass size: 640x640

lolicon

View Same Google ImgOps iqdb SauceNAO 23006-2835695622.jpg, 539KiB, 1408x828

lolicon Fri 19 May 2023 03:01:10 No.7383157 Report

Quoted By:

kiss kiss goodnight

Anonymous

Anonymous Fri 19 May 2023 03:09:53 No.7383163 Report

Quoted By: >>7383214

>>7383126
One gallery post still exists, so this guy has the lora. You can try commenting and asking him to upload it somewhere
https://civitai.com/posts/195328

Anonymous

View Same Google ImgOps iqdb SauceNAO upgrade.jpg, 109KiB, 640x640

Anonymous Fri 19 May 2023 03:12:42 No.7383165 Report

Quoted By: >>7383472

anyone know how to get teeth/smile like this for like a portrait

Anonymous

Anonymous Fri 19 May 2023 03:22:31 No.7383172 Report

Quoted By: >>7383173

>first dadaptation style train
>result is very true to the source and no frying
>however it keeps trying to spawn 2-5 women in each gen
what the deuce brian

Anonymous

Anonymous Fri 19 May 2023 03:23:32 No.7383173 Report

Quoted By: >>7383176

>>7383172
what setting /hyperparameters you use?

Anonymous

View Same Google ImgOps iqdb SauceNAO fullBodyComparison.jpg, 489KiB, 2045x2045

Anonymous Fri 19 May 2023 03:23:32 No.7383174 Report

Quoted By:

Was posting in the last thread, had a lot of people offering me advice, I had a lot of embedding spam. Thanks for the help to everyone who commented.

The new settings definitely make the lines sharper, The overall change is minor, but it's there. Here's an example of the same prompts with the settings changed:

old: https://files.catbox.moe/cwheqx.png
new: https://files.catbox.moe/me3df7.png

Only thing that's still getting me is full body shots. I've tried a few models, but I always look overcooked when I'm genning from head to toe. Only thing that's different between the images are some prompts to zoom in closer.

Full Body: https://files.catbox.moe/zxavuw.png
Closer: https://files.catbox.moe/xtkwm7.png

Anyone got advice for genning good full body pics, or are we doomed to just inpaint it away?

Anonymous

View Same Google ImgOps iqdb SauceNAO 00183-3187800857.png, 2MiB, 1024x1536

Anonymous Fri 19 May 2023 03:24:29 No.7383175 Report

Quoted By:

Anonymous

Anonymous Fri 19 May 2023 03:26:15 No.7383176 Report

Quoted By:

>>7383173
i was about to say nothing unusual, but i just realized i left my learning rates on adamw settings instead of changing them to 1.0

Anonymous

Anonymous Fri 19 May 2023 03:31:00 No.7383182 Report

Quoted By:

>>7382809
Dang, catbox for them boobas?

Anonymous

Anonymous Fri 19 May 2023 03:36:04 No.7383188 Report

Quoted By: >>7383336

>>7383015
>overfit
how do you know that's not just the art style of the lora?

Anonymous

Anonymous Fri 19 May 2023 03:37:42 No.7383190 Report

Quoted By:

Is there any integrated way in AUTOMATIC1111 to make a gif, when doing a loopback img2img?

Anonymous

Anonymous Fri 19 May 2023 03:39:36 No.7383191 Report

Quoted By:

Sometimes I get really fucking terrible results, like deviantart teenage autist drawing-bad. And I don't undetstand why, the prompts are similar to other gens where I've had great results.
It's like that one or two prompts tapped into some really specific shitty artist.

Anyone else gets the same?

Anonymous

Anonymous Fri 19 May 2023 03:40:55 No.7383194 Report

Quoted By: >>7383198 >>7383203

what fucking upscaler do I use to not look all wonky

Anonymous

View Same Google ImgOps iqdb SauceNAO 05664-3040637421-nsfw, masterpie (...).png, 940KiB, 768x1152

Anonymous Fri 19 May 2023 03:49:04 No.7383198 Report

Quoted By: >>7383203 >>7383230

>>7383194
I use remacri, you have to download it

Anonymous

Anonymous Fri 19 May 2023 03:56:10 No.7383202 Report

Quoted By: >>7383307

>>7383045
>>7383049
Sorry for the dreail but can you do those without the robot lines?

Anonymous

Anonymous Fri 19 May 2023 03:56:32 No.7383203 Report

Quoted By: >>7383230

>>7383194
Check out people's catbox metadata. Like >>7383198 said, I also use Remacri.
https://upscale.wiki/wiki/Model_Database

Anonymous

Anonymous Fri 19 May 2023 04:03:01 No.7383210 Report

Quoted By: >>7383541

>>7382567
when i think ive caught up, im reminded you lapped me, how do you construct such beauty? how can i replicate this pose in my foolish, knock-off inspired works?

Anonymous

Anonymous Fri 19 May 2023 04:04:35 No.7383211 Report

Quoted By: >>7383541

>>7382746
you're just too good...

Anonymous

View Same Google ImgOps iqdb SauceNAO 00289-3875260869.png, 921KiB, 768x1024

Anonymous Fri 19 May 2023 04:09:39 No.7383213 Report

Quoted By: >>7383368

Used a wedgie LoRA to get this, but it wasn't really made for this; more often than not it's the girl pulling her panties (or leotard) to give herself a wedgie. Anyone have advice on how to get clothes between labia more consistently (without the self inflicted wedgies)?

Anonymous

Anonymous Fri 19 May 2023 04:12:16 No.7383214 Report

Quoted By:

>>7383163
I found it, thanks though.
anon files
/2au3e8sbz5/Lakilolom_style_v1_zip

Anonymous

Anonymous Fri 19 May 2023 04:13:03 No.7383215 Report

Quoted By:

How should I upscale images when preparing a dataset? img2img at low denoise? some separate upscaler program?

Anonymous

View Same Google ImgOps iqdb SauceNAO 1682788819222.png, 744KiB, 512x768

Anonymous Fri 19 May 2023 04:18:53 No.7383221 Report

Quoted By:

>My gen got used for OP
Nice, thanks OP.

Anonymous

View Same Google ImgOps iqdb SauceNAO aftersexslut.jpg, 840KiB, 808x1200

Anonymous Fri 19 May 2023 04:25:39 No.7383230 Report

Quoted By: >>7383244 >>7383252

>>7383203
>>7383198
Anons... how do I get this thing to be usable in the webui? It's downloaded, it's in the folder, it shows up under extensions, but it's not an option for my upscaling

some /h/entai for your time

Anonymous

View Same Google ImgOps iqdb SauceNAO tmpuuroa6vw.png, 2MiB, 1024x1536

Anonymous Fri 19 May 2023 04:30:05 No.7383242 Report

Quoted By: >>7383263

Anonymous

View Same Google ImgOps iqdb SauceNAO 00458-1723895434.png, 3MiB, 1024x2048

Anonymous Fri 19 May 2023 04:33:17 No.7383244 Report

Quoted By: >>7383248

>>7383230
Have you restarted the webui since downloading it? You can't just refresh the ui, you have to shut it down and start it up again.

Anonymous

Anonymous Fri 19 May 2023 04:35:56 No.7383248 Report

Quoted By: >>7383249 >>7383250

>>7383244
>>7383244
Yes, still not showing up. Is it supposed to be automatic?

Anonymous

Anonymous Fri 19 May 2023 04:37:14 No.7383249 Report

Quoted By: >>7383255

>>7383248
Yes.
You put the .pth in .../stable-diffusion-webui/models/ESRGAN/ ?

Anonymous

Anonymous Fri 19 May 2023 04:39:02 No.7383250 Report

Quoted By:

>>7383248
If it's in txt2img you want it, you have to add a line in the ui settings. Something something add upscaler.
You will get a drop down menu to select upscalers, like how you can choose clipskip.

Anonymous

Anonymous Fri 19 May 2023 04:40:17 No.7383252 Report

Quoted By: >>7383255

>>7383230
is it in your esrgan folder

Anonymous

Anonymous Fri 19 May 2023 04:41:52 No.7383254 Report

Quoted By: >>7383399 >>7383405

Is there a way to slow down your GPU when prompting? I don't like the idea of my new GPU going as hard as it can overnight, every night.

Anonymous

View Same Google ImgOps iqdb SauceNAO creamie.png, 1MiB, 1080x1542

Anonymous Fri 19 May 2023 04:42:03 No.7383255 Report

Quoted By: >>7383436

>>7383249
>>7383252
That was it, it was in extensions folder, thanks. Populated some checkboxes that didn't do anything

Anonymous

Anonymous Fri 19 May 2023 04:54:05 No.7383263 Report

Quoted By:

>>7383242
awesome

Anonymous

View Same Google ImgOps iqdb SauceNAO 00165.png, 2MiB, 1536x3072

Anonymous Fri 19 May 2023 04:54:26 No.7383266 Report

Quoted By: >>7383271 >>7383642

It was not my intention to create this but... rocket launch girls sounds like a nice idea now that I see it. Wat do?

Anonymous

Anonymous Fri 19 May 2023 04:58:19 No.7383269 Report

Quoted By: >>7383277

I am making a single character lora. Is it advisable to completely avoid using images with other chars, photoshop/crop them out as much as possible or use them as is with proper tagging?

Anonymous

View Same Google ImgOps iqdb SauceNAO 00460-2556415833.png, 2MiB, 1024x1536

Anonymous Fri 19 May 2023 04:58:24 No.7383270 Report

Quoted By:

Anonymous

Anonymous Fri 19 May 2023 04:59:33 No.7383271 Report

Quoted By: >>7383376

>>7383266
Is it a tag combo you stumbled across or did you accidentally make it with a LoRA?

Anonymous

Anonymous Fri 19 May 2023 05:01:23 No.7383273 Report

Quoted By: >>7383282 >>7383284

>>7382864
as much as this nigger's ass is dooming over shit he has a point on
>nude version of sdg; 1girl, smile

Anonymous

Anonymous Fri 19 May 2023 05:03:49 No.7383277 Report

Quoted By: >>7383328

>>7383269
Depends how many of them

Anonymous

Anonymous Fri 19 May 2023 05:05:18 No.7383281 Report

Quoted By:

>>7382726
remove the AMD link it's wrong

Anonymous

Anonymous Fri 19 May 2023 05:05:37 No.7383282 Report

Quoted By: >>7383288

>>7383273
What else would you want to see itt, barely anyone puts effort into porn

Anonymous

View Same Google ImgOps iqdb SauceNAO 00462-2535944401.png, 2MiB, 1024x1536

Anonymous Fri 19 May 2023 05:07:29 No.7383284 Report

Quoted By: >>7383288

Really loving these magical girl prompts. I always loved the skirt on the sides with the front exposed.

Before fixing face with inpaint (trying out the pussy peek LoRA to get a really low cut leotard, but it's not happening).
https://files.catbox.moe/0qqa7u.png

>>7383273
Excuse you, mine is very different. She's not nude.
In general, I post to the board that I'm allowed. If it's even a little /d/eviant, I put it in /d. If it's safe, I put it in /g.

Anonymous

Anonymous Fri 19 May 2023 05:10:04 No.7383288 Report

Quoted By:

>>7383282
at least some (flat color) and (sketch) instead of 2.5d stuff thanks to (masterpiece), (best quality)

>>7383284
>mine is very different. She's not nude.
based

Anonymous

Anonymous Fri 19 May 2023 05:13:04 No.7383293 Report

Quoted By: >>7383296

I started making a new guide for the OP, but I stopped when I realized that I get nothing for spoonfeeding newfags
and I also hate most of you

Anonymous

Anonymous Fri 19 May 2023 05:17:17 No.7383296 Report

Quoted By:

>>7383293
Thanks for telling us anon. I'm sure we'll dearly miss the wisdom you could have provided.

Anonymous

View Same Google ImgOps iqdb SauceNAO paperspace.png, 3KiB, 269x69

Anonymous Fri 19 May 2023 05:23:41 No.7383303 Report

Quoted By: >>7383309

Is there anything I can do with this to help people out?

Anonymous

View Same Google ImgOps iqdb SauceNAO 00820-2865917713.png, 2MiB, 1024x1536

Anonymous Fri 19 May 2023 05:27:04 No.7383307 Report

Quoted By:

>>7383202
Naah, I've posted a lot of other pictures of regular humans in the cowgirl position already.

Anonymous

Anonymous Fri 19 May 2023 05:27:28 No.7383309 Report

Quoted By:

>>7383303
create a preview image for every lora in the gayshit

Anonymous

View Same Google ImgOps iqdb SauceNAO 108252375_p1_master1200.jpg, 847KiB, 800x1200

Anonymous Fri 19 May 2023 05:33:15 No.7383318 Report

Quoted By: >>7383319 >>7383322 >>7383326 >>7383373

any /h/ anons have any idea what model/lora is being used to generate this face? from ape_loo on cuckxiv.

Anonymous

Anonymous Fri 19 May 2023 05:34:29 No.7383319 Report

Quoted By: >>7383342

>>7383318
nice ad

Anonymous

Anonymous Fri 19 May 2023 05:38:01 No.7383322 Report

Quoted By:

>>7383318
Yes, out of thousands of them I know the one. I know it because he inpainted it. No i will not tell you. Yes you should buy a rope

Anonymous

Anonymous Fri 19 May 2023 05:40:07 No.7383325 Report

Quoted By:

>4060ti
>$500
>128bit bus
poorfags continue to lose lmao

Anonymous

Anonymous Fri 19 May 2023 05:40:11 No.7383326 Report

Quoted By: >>7383335

>>7383318
I'm more distracted by the ass-tier work done on the stitching from the collar.

Anonymous

Anonymous Fri 19 May 2023 05:40:23 No.7383328 Report

Quoted By: >>7383362

>>7383277
Currently 34 out of 375 aren't (1girl), yet I can photoshop some of those to make them solo and put less repeats on them in general.

Another question - how much training is too much? At what number of overall repeats should I aim to not strain my GPU on overcooking LoRA?

Anonymous

Anonymous Fri 19 May 2023 05:43:25 No.7383335 Report

Quoted By: >>7383342 >>7383345

>>7383326
yeah everything is mostly all shit. u can tell waht's being imprinted lmao

Anonymous

Anonymous Fri 19 May 2023 05:43:29 No.7383336 Report

Quoted By:

>>7383188
because it's verbatim jinx's hair but red. i bet the prompt use didn't even specify a hairstyle

Anonymous

Anonymous Fri 19 May 2023 05:47:42 No.7383342 Report

Quoted By: >>7383347

>>7383335
I'm not sure I agree with it all being shit. It's a good picture, but the stitching is the first thing that caught my eye.

Also I just wanted to dunk on the retarded question. We'd need more info than just one picture to figure out a model, and >>7383319 is right. It reeks of an ad.

Anonymous

View Same Google ImgOps iqdb SauceNAO 0403b51d25ee83d5ffed1e08b610487f.jpg, 372KiB, 862x1027

Anonymous Fri 19 May 2023 05:49:01 No.7383344 Report

Quoted By: >>7383347

Was there an update overnight or something?
Everything I do suddenly looks like absolute trash like this.

Anonymous

Anonymous Fri 19 May 2023 05:50:11 No.7383345 Report

Quoted By:

>>7383335
Personally, if I'm doing moderate-heavy inpainting, I do a final, full image img2img at very low denoising to iron everything out. Do people not do that?

Anonymous

Anonymous Fri 19 May 2023 05:52:36 No.7383347 Report

Quoted By: >>7383356

>>7383342
it screams sakimichan wannabe style, of course its shit

>>7383344
>pulls after each update
kek

if you're using xformers try reloading webui, sometimes because of it ALL your further gens will be shit

Anonymous

Anonymous Fri 19 May 2023 06:00:35 No.7383356 Report

Quoted By:

>>7383347
I don't manually update, but sometimes I see stuff updating in the launch cmd.
Restarting a1111 entirely seemed to have fixed it though.

Anonymous

Anonymous Fri 19 May 2023 06:05:22 No.7383362 Report

Quoted By:

>>7383328
If it's an easy crop that won't leave weird background then crop. If it's not but it's a cool picture you wanna keep then lower repeats. Otherwise nuke it.
>How much is too much
When it's too much it's too much. Depends on everything else, just save every 1 or 2 epochs.

Anonymous

Anonymous Fri 19 May 2023 06:12:55 No.7383368 Report

Quoted By:

>>7383213
try drawing a very thin inpainting mask because the lips and see how that fairs with a promp of "thong" or something of that sort

Anonymous

View Same Google ImgOps iqdb SauceNAO 00219-3016667979.png, 2MiB, 1024x1536

Anonymous Fri 19 May 2023 06:13:41 No.7383369 Report

Quoted By:

Anonymous

Anonymous Fri 19 May 2023 06:15:54 No.7383373 Report

Quoted By: >>7383392 >>7383425

>>7383318
>took the time out to fix the face
>doesn't bother fixing continuity error in background
why even bother?

Anonymous

Anonymous Fri 19 May 2023 06:21:01 No.7383376 Report

Quoted By:

>>7383271
>https://gitgud.io/gayshit/makesomefuckingporn
Just downloaded a Videl lora (videl_v10) right before this and was messing around with it. Just a retarded semen everywhere prompt. Now that I've seen it though, I gotta figure out how to reproduce it but without it being a low quality shit image.

Anonymous

Anonymous Fri 19 May 2023 06:21:25 No.7383377 Report

Quoted By: >>7383392

These threads have made me appreciate anons who invest time and effort in their creations when it seems like everywhere else people are producing poorly made images in bulk.

Why bother if the final result doesn't even resemble the character?

Anonymous

Anonymous Fri 19 May 2023 06:35:54 No.7383392 Report

Quoted By: >>7383396

>>7383373
why even bother prompting some sick ass (detailed background:1.4) that would be nearly impossible to fix

its time to become (simple background) chad

>>7383377
i've skimmed through ~300 AI gens of Hiiragi Kagami on pixiv and i couldn't even recognize her there except for TWO of them that retained her eye shape and weren't made through LuckyStar style lora

i don't mind pixiv banning AI gens because its fucking drowning in low quality stuff atm

Anonymous

Anonymous Fri 19 May 2023 06:38:57 No.7383396 Report

Quoted By: >>7383405 >>7383427

>>7383392
>that would be nearly impossible to fix
the background in the chun-li image the shill posted i could fix in photoshop in 30 seconds tops

Anonymous

Anonymous Fri 19 May 2023 06:41:12 No.7383399 Report

Quoted By:

>>7383254
Yes

Anonymous

Anonymous Fri 19 May 2023 06:44:22 No.7383405 Report

Quoted By: >>7383412

>>7383396
you're implying that average sakimichan enjoyer knows how to use photoshop

>>7383254
try undervolting it through afterburner

Anonymous

Anonymous Fri 19 May 2023 06:47:42 No.7383412 Report

Quoted By:

>>7383405
>you're implying that average sakimichan enjoyer knows how to use photoshop
kek

Anonymous

View Same Google ImgOps iqdb SauceNAO xyz_grid-0018-1937326473.png, 4MiB, 6656x589

Anonymous Fri 19 May 2023 06:52:50 No.7383414 Report

Quoted By: >>7383423

Is there a benefit to having more steps than necessary to make the image look good? What's the reason for doing 20-30 steps when even 13-15 looks good?

Anonymous

Anonymous Fri 19 May 2023 07:02:36 No.7383423 Report

Quoted By:

>>7383414
Depends if your sampler converges or not.

Anonymous

Anonymous Fri 19 May 2023 07:03:15 No.7383425 Report

Quoted By: >>7383432

>>7383373
Why fix everything about an image? It'd take ages in some cases. For me it's fine if I achieve the general thing I wanted, I'm not gonna go around inpainting all details

Anonymous

Anonymous Fri 19 May 2023 07:05:08 No.7383427 Report

Quoted By: >>7383432

>>7383396
How would you even Photoshop that?

Anonymous

Anonymous Fri 19 May 2023 07:12:03 No.7383432 Report

Quoted By: >>7383453 >>7383455

>>7383425
because it just looks half-assed if you don't? there isn't even much to fix with the chun-li one; the bridge on the right side of the photo just needs to be lowered and it's basically good
>>7383427
>select subject, feather, inverse mask, clone stamp one side of the bridge and lower it such that it's properly lined up
>select subject, feather, duplicate layer, mask, work on the bottom layer and do as above; either clone stamp or just select one side of the bridge and move it down
after that, it's just a matter of spot healing/content aware fill, and if you do a shitty job, inpaint at low denoise will blend it anyways.
there's not one way to do it, but it's a very simple fix that would take barely any time

Anonymous

Anonymous Fri 19 May 2023 07:14:25 No.7383433 Report

Quoted By: >>7383544

>>7382638
Chocobox?

Anonymous

Anonymous Fri 19 May 2023 07:20:20 No.7383436 Report

Quoted By:

>>7383255
Catbox?

Anonymous

View Same Google ImgOps iqdb SauceNAO image_2023-05-19_033421402.png, 686KiB, 768x768

Anonymous Fri 19 May 2023 07:34:25 No.7383444 Report

Quoted By: >>7383727 >>7384583

Anonymous

Anonymous Fri 19 May 2023 07:46:00 No.7383453 Report

Quoted By: >>7383461 >>7383561

>>7383432
>30 seconds
More like 15 minutes at least

Anonymous

Anonymous Fri 19 May 2023 07:47:42 No.7383455 Report

Quoted By: >>7383461 >>7383559

>>7383432
>half-assed
Yeah and? It's AI, it conveys the main girl just fine, what else would you need? It's not some art piece to be rated but an equivalent of a shitty sketch by a beginner artist, you don't go comment on issues in some doodles do you?

Anonymous

View Same Google ImgOps iqdb SauceNAO WAKEMEUP.jpg, 8KiB, 589x40

Anonymous Fri 19 May 2023 07:49:25 No.7383459 Report

Quoted By: >>7383464 >>7383471 >>7383477 >>7383483 >>7383484

>testing a shit ton of style loras from gayshit and civit
>almost all of them are overfit
>trying to figure out if it's my settings or the model
>go to one model's page
>picrel at the very bottom
I know this has been posted a bunch before but I'm so sick of this meme. Train your shit right and test the generalization, not just how well it can create images exactly like the dataset. Also, if I see anyone posting test grids here at less than weight 1 I'm going to personally fucking punt you into the sun since you love fried shit so much.

Anonymous

Anonymous Fri 19 May 2023 07:51:45 No.7383461 Report

Quoted By: >>7383480

>>7383453
>>7383455
hehe 1girl, smile go BRRRRRRR

Anonymous

Anonymous Fri 19 May 2023 07:54:15 No.7383463 Report

Quoted By:

Anyone come across this error when trying regional prompter?

RuntimeError: Expected query.size(0) == key.size(0) to be true, but got false.

Worked the other day.

Anonymous

Anonymous Fri 19 May 2023 07:55:34 No.7383464 Report

Quoted By: >>7383467

>>7383459
make your own style loras nigga

Anonymous

Anonymous Fri 19 May 2023 07:56:07 No.7383465 Report

Quoted By: >>7383773

>>7382578
can you at least roughly explain how this is done..

Anonymous

Anonymous Fri 19 May 2023 07:57:44 No.7383467 Report

Quoted By: >>7383474

>>7383464
anon did you know that you can use other people's loras as well as the ones you train yourself?

Anonymous

Anonymous Fri 19 May 2023 07:59:34 No.7383469 Report

Quoted By: >>7383471

Does anyone have the alexi lora that got taken down or is everyone using the civit one?

Anonymous

Anonymous Fri 19 May 2023 08:00:45 No.7383471 Report

Quoted By: >>7383476

>>7383459
adding to this, holy shit some of your tags are fucking atrocious. Who doesn't at least go through to remove shit like "commentary request". Are you people seriously just training with the tags you get from grabber?

>>7383469
fuck off glowie it's literally everywhere

Anonymous

Anonymous Fri 19 May 2023 08:01:51 No.7383472 Report

Quoted By: >>7383743

>>7383165
hiramoto anon?

Anonymous

Anonymous Fri 19 May 2023 08:02:39 No.7383474 Report

Quoted By:

>>7383467
of fucking course, and if you don't like the quality of stuff you're getting FOR FREE then start training loras yourself nigga

Anonymous

View Same Google ImgOps iqdb SauceNAO 00009-3492718828.png, 2MiB, 1024x1536

Anonymous Fri 19 May 2023 08:03:20 No.7383475 Report

Quoted By:

>mfw I'm thirsty and want some choccy milk

Anonymous

Anonymous Fri 19 May 2023 08:04:24 No.7383476 Report

Quoted By:

>>7383471
Found it, thanks

Anonymous

Anonymous Fri 19 May 2023 08:04:26 No.7383477 Report

Quoted By: >>7383510

>>7383459
Loras that work at 1 are underbaked shit. 0.95 is peak, brainlet, skirting the edge.
>not testing style falloff to see how well it produces in-between
lol
lmao
Learn how to train

Anonymous

Anonymous Fri 19 May 2023 08:04:46 No.7383478 Report

Quoted By:

Whats better for 2girls and more regional prompter, multidiffusion or just good old regular gacha?

Anonymous

Anonymous Fri 19 May 2023 08:06:03 No.7383480 Report

Quoted By:

>>7383461
Unironically this. If you're telling me there's an issue with "1girl, spread legs, smile" your opinion is automatically invalid

Anonymous

Anonymous Fri 19 May 2023 08:08:45 No.7383483 Report

Quoted By:

>>7383459
NOOOOO DONT MAKE ME LOWER THE SLIDER BELOW 1 NOOOO MY FRAGILE 100% MENTALITY AHHHHH
>meanwhile it works just fine at 0.5 and all anon had to do was lower slider
Truly tragic.

Anonymous

Anonymous Fri 19 May 2023 08:10:33 No.7383484 Report

Quoted By: >>7383496 >>7383515

>>7383459
On this topic, what if you already have a bunch of loras and don't want to overflow? Would an overfitted lora at 0.5 and a good lora at 1.0 cause the same issue if I have a handful of loras already enabled?

Anonymous

Anonymous Fri 19 May 2023 08:10:37 No.7383485 Report

Quoted By:

some weak bait this thread

Anonymous

Anonymous Fri 19 May 2023 08:21:10 No.7383496 Report

Quoted By:

>>7383484
There shouldn't be any issues if the overfit lora actually works properly at some strength

Anonymous

Anonymous Fri 19 May 2023 08:22:41 No.7383498 Report

Quoted By: >>7383500

Is there a 'opening your front door and seeing an anime girl there' lora or should I add that to the list of stuff to make?

Anonymous

Anonymous Fri 19 May 2023 08:26:25 No.7383500 Report

Quoted By: >>7383503

>>7383498
exactly what we need, yes, another fried LoRA for something easily promptable

Anonymous

Anonymous Fri 19 May 2023 08:28:30 No.7383503 Report

Quoted By:

>>7383500
Hold your anticipation, anon.

Anonymous

Anonymous Fri 19 May 2023 08:31:25 No.7383510 Report

Quoted By: >>7383516

>>7383477
what are some of your loras

Anonymous

Anonymous Fri 19 May 2023 08:37:42 No.7383512 Report

Quoted By: >>7383516

I wonder how well training a general purpose LoRA with a huge dataset on 2048x for better anatomy on higher res base gens would go.

Anonymous

View Same Google ImgOps iqdb SauceNAO 00181-781039555.png, 1MiB, 1024x1536

Anonymous Fri 19 May 2023 08:40:13 No.7383515 Report

Quoted By:

>>7383484
I've run into this scenario a bunch when style mixing. The overfit model poisons the well and you have to lower the weights so low that it barely registers. Of course even when you do that it'll still manage fuck up the details and the background

Anonymous

Anonymous Fri 19 May 2023 08:40:29 No.7383516 Report

Quoted By: >>7383518 >>7383766

>>7383512
Isn't this how AOM was made?

>>7383510
How is this relevant, I'm not talking about training settings.

Anonymous

Anonymous Fri 19 May 2023 08:41:26 No.7383517 Report

Quoted By: >>7383521

it's all so tiresome...

Anonymous

Anonymous Fri 19 May 2023 08:41:54 No.7383518 Report

Quoted By: >>7383533

>>7383516
No.

Anonymous

View Same Google ImgOps iqdb SauceNAO 00133.png, 2MiB, 1152x1440

Anonymous Fri 19 May 2023 08:45:11 No.7383521 Report

Quoted By: >>7383532

>>7383517
what is?

Anonymous

Anonymous Fri 19 May 2023 08:46:14 No.7383522 Report

Quoted By: >>7383525

been gone for a month, anything cool I missed?

Anonymous

Anonymous Fri 19 May 2023 08:47:32 No.7383525 Report

Quoted By:

>>7383522
a steep decline in image quality, if that counts

Anonymous

Anonymous Fri 19 May 2023 08:51:13 No.7383532 Report

Quoted By: >>7383536 >>7383770

>>7383521
for asking peoples catbox but they dont want to give it.
catbox?

Anonymous

Anonymous Fri 19 May 2023 08:51:33 No.7383533 Report

Quoted By: >>7383535

>>7383518
How was AOM made then, iirc it was trained on Instagram models to improve bodies? Or do you only mean the resolution difference

Anonymous

Anonymous Fri 19 May 2023 08:54:43 No.7383535 Report

Quoted By:

>>7383533
I suspected you were retarded, but this confirms it

Anonymous

View Same Google ImgOps iqdb SauceNAO 00132a.png, 1MiB, 1075x1344

Anonymous Fri 19 May 2023 08:55:03 No.7383536 Report

Quoted By: >>7383920

>>7383532
odd thing to be tired of
https://files.catbox.moe/1hwzfk.png
I think (disgust) was a wildcard of face expressions, not like it matters

Anonymous

Anonymous Fri 19 May 2023 08:59:03 No.7383540 Report

Quoted By:

Any tips for generating older looking women? I try using (mature woman) and (beautiful face) but they rarely look old enough.

Anonymous

View Same Google ImgOps iqdb SauceNAO 9996490989765788-1417683485-kill (...).png, 3MiB, 2400x1600

Anonymous Fri 19 May 2023 09:00:43 No.7383541 Report

Quoted By: >>7383558 >>7383732

>>7383210
just look peoples catboxes and just edit them for you liking. inpaint just takes most of my killjoy work. if you want my catbox heres original one
https://files.catbox.moe/7mi05m.png
>>7383211
thank you. i apprentice it

Anonymous

View Same Google ImgOps iqdb SauceNAO 00124-2471334675-nsfw, scan, ((( (...).png, 2MiB, 1024x1536

Anonymous Fri 19 May 2023 09:08:10 No.7383544 Report

Quoted By:

>>7383433
I love camisoles.
https://files.catbox.moe/i29j0r.png

Anonymous

Anonymous Fri 19 May 2023 09:24:41 No.7383558 Report

Quoted By:

>>7383541
boob bag

Anonymous

Anonymous Fri 19 May 2023 09:24:42 No.7383559 Report

Quoted By: >>7383560

>>7383455
>what else would you need?
for the background to either not exist or be coherent. it's not that hard, not that complicated, and you're just lazy

Anonymous

Anonymous Fri 19 May 2023 09:25:32 No.7383560 Report

Quoted By: >>7383563

>>7383559
thats just being autistic

Anonymous

Anonymous Fri 19 May 2023 09:25:43 No.7383561 Report

Quoted By: >>7383564

>>7383453
>15 minutes
lmao it takes you 15 minutes to magic wand and click a few buttons, then alt+left click. no lol. you literally don't have a clue

Anonymous

Anonymous Fri 19 May 2023 09:26:48 No.7383563 Report

Quoted By: >>7383566

>>7383560
this is why pixiv and other retarded sites are banning ai sites, because low effort retards that can't be bothered to make a coherent image when it's handed to them on a platter spam the sites with their garbage

Anonymous

Anonymous Fri 19 May 2023 09:27:14 No.7383564 Report

Quoted By: >>7383565

>>7383561
it sounds simple when you type but takes longer when you actually have to do it

Anonymous

Anonymous Fri 19 May 2023 09:27:51 No.7383565 Report

Quoted By: >>7383569

>>7383564
how would you know? you've never done it and you're just shitposting

Anonymous

View Same Google ImgOps iqdb SauceNAO file.png, 1MiB, 858x1200

Anonymous Fri 19 May 2023 09:28:27 No.7383566 Report

Quoted By: >>7383567 >>7383568

>>7383563
ok so why are they allowing doodles like pic related? shouldnt they be banned too?

Anonymous

Anonymous Fri 19 May 2023 09:29:01 No.7383567 Report

Quoted By:

>>7383566
too much SOVL

Anonymous

Anonymous Fri 19 May 2023 09:29:30 No.7383568 Report

Quoted By: >>7383571

>>7383566
because there is an growing, infinitely larger supply of shitty ai images that will basically take the role of pic related but tenfold

Anonymous

Anonymous Fri 19 May 2023 09:29:32 No.7383569 Report

Quoted By: >>7383573

>>7383565
because im not retarded and I know lining up the fucking image via photoshop wouldnt be a fun time. Normally you'd just select all of it and inpaint like a human being without playing with shop

Anonymous

Anonymous Fri 19 May 2023 09:30:47 No.7383571 Report

Quoted By: >>7383577 >>7383584

>>7383568
AI isnt that bad though, you're complaining about a background detail lol

Anonymous

Anonymous Fri 19 May 2023 09:32:05 No.7383573 Report

Quoted By: >>7383576

>>7383569
>lining up the fucking image
what are you talking about. you click on the source for the clonestamp, drag the mouse over to approximately where you want it to be stamped, and click. that's all there is to it. just google "clone stamp" on youtube if you're this fucking stupid

Anonymous

View Same Google ImgOps iqdb SauceNAO tmpzuw6hutl.png, 2MiB, 1536x1024

Anonymous Fri 19 May 2023 09:32:20 No.7383574 Report

Quoted By: >>7383575 >>7383578

Anonymous

Anonymous Fri 19 May 2023 09:34:56 No.7383575 Report

Quoted By:

>>7383574
Nice clussy

Anonymous

Anonymous Fri 19 May 2023 09:37:51 No.7383576 Report

Quoted By: >>7383579

>>7383573
doesnt seem like it would work for that background without issues

Anonymous

Anonymous Fri 19 May 2023 09:37:59 No.7383577 Report

Quoted By:

>>7383571
>AI isnt that bad though
newfag alert

Anonymous

Anonymous Fri 19 May 2023 09:38:28 No.7383578 Report

Quoted By: >>7383586

>>7383574
catbox?

Anonymous

Anonymous Fri 19 May 2023 09:38:43 No.7383579 Report

Quoted By:

>>7383576
i'm bored already. you need to put more effort into your bait

Anonymous

Anonymous Fri 19 May 2023 09:40:54 No.7383581 Report

Quoted By: >>7383584

What is the right way to use a tag like marker_(medium)? I know the underscore is unnecessary so would it be "marker (medium)" or is the parenthesis also not needed?

Anonymous

Anonymous Fri 19 May 2023 09:42:31 No.7383584 Report

Quoted By: >>7383585 >>7383633

>>7383571
>AI isnt that bad though
lmao, lol. i've been an avid user ever since voldy's webui came into existence, and this is just flat out bullshit. the influx of shitty, low effort sameface nai gens, half of which probably lack a vae, is exactly why pixiv is getting rid of ai art
>>7383581
>marker /(medium/)
escape brackets stops the parantheses from counting as tokens

Anonymous

Anonymous Fri 19 May 2023 09:44:01 No.7383585 Report

Quoted By: >>7383587

>>7383584
>vaeless
yeah ok my brain filtered them, fine.
>pixiv
?? they're banning ways to profit off it, NAI gens have nothing to do with this

Anonymous

Anonymous Fri 19 May 2023 09:45:07 No.7383586 Report

Quoted By: >>7385084

>>7383578
not the same image but the same parameters
https://files.catbox.moe/98gstj.png

Anonymous

Anonymous Fri 19 May 2023 09:48:53 No.7383587 Report

Quoted By: >>7383593 >>7383602

>>7383585
>?? they're banning ways to profit off it, NAI gens have nothing to do with this
my bad, i was thinking of other sites and conflated pixiv doing their shit with that. regardless, point stands. lazy ai gens are going to replace the shitty sea of poorly drawn hand drawn art that exists on all image hosting sites

Anonymous

Anonymous Fri 19 May 2023 09:54:16 No.7383593 Report

Quoted By:

>>7383587
They are, yeah. And that is part of why bothering with every detail on AI images is pointless - the more you do it the more you find and then you realize you've sunk 5 hours hunting hair pixels because the look like shit. And none of it matters if the main portion of the image looks fine - going back to the image that sparked it only the clothes needed actual inpainting, cause no one gives a shit about generic backgrounds being misaligned.

Anonymous

View Same Google ImgOps iqdb SauceNAO 00174-2102794408.png, 1MiB, 768x1024

Anonymous Fri 19 May 2023 09:56:58 No.7383596 Report

Quoted By: >>7383635

Anonymous

Anonymous Fri 19 May 2023 10:03:42 No.7383602 Report

Quoted By: >>7383739

>>7383587
>lazy ai gens are going to replace the shitty sea of poorly drawn hand drawn art

one poorly hand drawn piece of art takes same amount of time as generating (big boobah) prompt with batchsize of AT LEAST 10 images on medium-end GPU

it will not be a sea, it will be a fucking ocean of 512x512 nai sameface gens with large badonglers

Anonymous

Anonymous Fri 19 May 2023 10:09:15 No.7383605 Report

Quoted By:

>>7383014
it's gotten noticeably worse like it does every year

Anonymous

View Same Google ImgOps iqdb SauceNAO s_mikus.jpg, 529KiB, 1152x3072

Anonymous Fri 19 May 2023 10:40:53 No.7383622 Report

Quoted By: >>7383734

Anonymous

Anonymous Fri 19 May 2023 10:42:46 No.7383628 Report

Quoted By: >>7383634 >>7385121 >>7386032

Anonymous

Anonymous Fri 19 May 2023 10:54:48 No.7383633 Report

Quoted By: >>7384951

>>7383584
>months of usage
>escapes brackets wrong
ouch

Anonymous

Anonymous Fri 19 May 2023 10:55:44 No.7383634 Report

Quoted By:

>>7383628
>newfags visit here learn from us then decide new rules.
sad if true. newfags should adapt.

>>7382742
>>7382726
which reminds me, dont forget to list threads on other boards again

Anonymous

Anonymous Fri 19 May 2023 10:56:12 No.7383635 Report

Quoted By: >>7383653

>>7383596
This might be the first image that's not mine that I'll save.
Catbox?

Anonymous

View Same Google ImgOps iqdb SauceNAO catbox_h5nbox.png, 1MiB, 768x1152

Anonymous Fri 19 May 2023 10:59:49 No.7383640 Report

Quoted By:

not good at fixing hands yet, but this is one of my first ones that came out alright. used BloodborneAnons prompts and changed them up, my negatives are all fucked up from a style i forgot to remove but its whatever.. im a newfag to this so yea

Anonymous

Anonymous Fri 19 May 2023 11:01:27 No.7383642 Report

Quoted By:

>>7383266
>Gohan teaches Videl to fly (GONE SEXUAL)

Anonymous

Anonymous Fri 19 May 2023 11:04:26 No.7383646 Report

Quoted By: >>7383657

do people still mess around with LoHas and LoCons or should I just stick with LoRas?

Anonymous

Anonymous Fri 19 May 2023 11:05:25 No.7383649 Report

Quoted By: >>7383654

what a miserable thread

Anonymous

Anonymous Fri 19 May 2023 11:06:25 No.7383650 Report

Quoted By: >>7383679 >>7383699

>want to try dadaptation
>8gb vramlet
owari da...

Anonymous

Anonymous Fri 19 May 2023 11:09:17 No.7383653 Report

Quoted By:

>>7383635
here you go
https://files.catbox.moe/o36isq.png

Anonymous

Anonymous Fri 19 May 2023 11:14:53 No.7383654 Report

Quoted By: >>7383665

>>7383649
still better than blacked threads

Anonymous

Anonymous Fri 19 May 2023 11:17:00 No.7383657 Report

Quoted By: >>7383699 >>7383771

>>7383646
locons are just better than loras

Anonymous

Anonymous Fri 19 May 2023 11:26:41 No.7383665 Report

Quoted By: >>7383666 >>7383673

>>7383654
is it really

Anonymous

Anonymous Fri 19 May 2023 11:27:09 No.7383666 Report

Quoted By:

>>7383665
yes unequivocally

Anonymous

Anonymous Fri 19 May 2023 11:28:02 No.7383668 Report

Quoted By: >>7383687

If I'm making a character lora and an overwhelming amount of the drawings are made by the same artist, how do I avoid baking the artstyle into the lora? Do I just lower the rank and alpha values?

Anonymous

Anonymous Fri 19 May 2023 11:30:26 No.7383673 Report

Quoted By:

>>7383665
got ur nose

Anonymous

Anonymous Fri 19 May 2023 11:35:06 No.7383679 Report

Quoted By:

>>7383650
colab?

Anonymous

View Same Google ImgOps iqdb SauceNAO 00023-2698717730.png, 2MiB, 1536x1024

Anonymous Fri 19 May 2023 11:41:31 No.7383683 Report

Quoted By: >>7383684

Anonymous

Anonymous Fri 19 May 2023 11:43:27 No.7383684 Report

Quoted By:

>>7383683
coolio

Anonymous

Anonymous Fri 19 May 2023 11:45:40 No.7383687 Report

Quoted By:

>>7383668
more repeats on the ones that aint by that artist

Anonymous

View Same Google ImgOps iqdb SauceNAO 00252-262121520.png, 2MiB, 1536x1024

Anonymous Fri 19 May 2023 11:59:20 No.7383691 Report

Quoted By:

>>7382725
catbox?

Anonymous

Anonymous Fri 19 May 2023 12:11:10 No.7383699 Report

Quoted By: >>7383719

>>7383650
I still wanna know if easyscripts sets the dadapt stuff by default

>>7383657
proof?

Anonymous

Anonymous Fri 19 May 2023 12:43:45 No.7383717 Report

Quoted By: >>7383721 >>7383722

>>7382753
Just want to say you're an absolute chad for using the catbox userscript

Anonymous

Anonymous Fri 19 May 2023 12:44:48 No.7383718 Report

Quoted By:

>>7382760
Inb4 this becomes paypig general

Anonymous

Anonymous Fri 19 May 2023 12:45:20 No.7383719 Report

Quoted By:

>>7383699
I did a ton of tests in the last 3 months, but it would be nice to do some more scientific stuff, I should cook something up to show

Anonymous

View Same Google ImgOps iqdb SauceNAO catbox_ibioxq.png, 1MiB, 1024x1024

Anonymous Fri 19 May 2023 12:45:44 No.7383721 Report

Quoted By: >>7383864

>>7383717
Thanks

Anonymous

Anonymous Fri 19 May 2023 12:46:43 No.7383722 Report

Quoted By: >>7383740

>>7383717
Hate that it doesn't work for me. I get the green symbol but nothing when I right click even though there's a tooltip there telling me to do it.

Anonymous

Anonymous Fri 19 May 2023 12:47:12 No.7383724 Report

Quoted By:

What region prompter extension do you all use?

Anonymous

Anonymous Fri 19 May 2023 12:48:59 No.7383727 Report

Quoted By:

>>7383444
based

Anonymous

Anonymous Fri 19 May 2023 12:52:57 No.7383732 Report

Quoted By:

>>7383541
based killjoy fucker

Anonymous

Anonymous Fri 19 May 2023 12:56:22 No.7383734 Report

Quoted By:

>>7383622
based 4koma enjoyer

Anonymous

Anonymous Fri 19 May 2023 12:57:42 No.7383736 Report

Quoted By: >>7384217

breeding killjoy...

Anonymous

Anonymous Fri 19 May 2023 13:00:41 No.7383739 Report

Quoted By:

>>7383602
At least I can get off to big badonga nai gens

Anonymous

Anonymous Fri 19 May 2023 13:02:16 No.7383740 Report

Quoted By: >>7383741

>>7383722
Are you using the userscript version of 4chanx? The extension version doesn't really work very well with it.

Anonymous

Anonymous Fri 19 May 2023 13:04:31 No.7383741 Report

Quoted By: >>7383748

>>7383740
Yep, although 4chanx is on greasemonkey while catbox is on violentmonkey. Don't think it worked while both were on greasemonkey either

Anonymous

Anonymous Fri 19 May 2023 13:07:40 No.7383743 Report

Quoted By:

>>7383472
maybe, I just know it from the phonk song upgrade

Anonymous

Anonymous Fri 19 May 2023 13:11:43 No.7383748 Report

Quoted By:

>>7383741
I dumped both on good old fashioned tampermonkey and everything seems to be working okay

Anonymous

Anonymous Fri 19 May 2023 13:18:51 No.7383752 Report

Quoted By:

Never tried Remacri but it actually seems like it's even more detailed than animesharp, interesting

Anonymous

Anonymous Fri 19 May 2023 13:44:43 No.7383766 Report

Quoted By:

>>7383516
well you told him his 1 weight lora concept was shit, it'd be cool to see the loras you made so we know that what you say is true

Anonymous

Anonymous Fri 19 May 2023 13:47:28 No.7383770 Report

Quoted By:

>>7383532
no one shares anything anymore and I liked checking out the thread for new info and how anons helped each other figure out new extensions and shit.

Anonymous

Anonymous Fri 19 May 2023 13:48:39 No.7383771 Report

Quoted By:

>>7383657
yeah no

Anonymous

View Same Google ImgOps iqdb SauceNAO kissanlaatikko.jpg, 3MiB, 2864x1072

Anonymous Fri 19 May 2023 13:49:41 No.7383773 Report

Quoted By: >>7384002

>>7383465
https://files.catbox.moe/crcee0.png
https://files.catbox.moe/7nlsbz.png

Anonymous

View Same Google ImgOps iqdb SauceNAO s_xyz_grid-0969-1819426649-full (...).jpg, 3MiB, 4416x4757

Anonymous Fri 19 May 2023 13:51:32 No.7383774 Report

Quoted By:

add detail lora is interesting, positive seems to ignore prompt quite a bit as you might imagine. negative is pretty nice

Anonymous

View Same Google ImgOps iqdb SauceNAO 00420-2103222886.png, 1MiB, 1024x1024

Anonymous Fri 19 May 2023 13:52:35 No.7383777 Report

Quoted By: >>7383780 >>7383913 >>7383967

>"j-just one peek, okay?"

Anonymous

Anonymous Fri 19 May 2023 13:58:54 No.7383780 Report

Quoted By: >>7383782

>>7383777
"Wow! They make me totally forget about your fingers, babe!"

Defiling 80s anime style with all the possible depravity we can think of today warms my heart.

Anonymous

View Same Google ImgOps iqdb SauceNAO 00006-2218069605.png, 1MiB, 1024x1536

Anonymous Fri 19 May 2023 14:01:15 No.7383781 Report

Quoted By:

Anonymous

Anonymous Fri 19 May 2023 14:01:42 No.7383782 Report

Quoted By: >>7383798

>>7383780
kek I actually posted the version without inpainting by mistake, how silly of me

>Defiling 80s anime style with all the possible depravity we can think of today warms my heart.
just think of the possibilities!

Anonymous

View Same Google ImgOps iqdb SauceNAO s_xyz_grid-0970-2127615595-full (...).jpg, 3MiB, 4416x4757

Anonymous Fri 19 May 2023 14:07:29 No.7383790 Report

Quoted By: >>7383805

freaky

Anonymous

View Same Google ImgOps iqdb SauceNAO 00047-1264293370.png, 1MiB, 1024x1024

Anonymous Fri 19 May 2023 14:23:12 No.7383795 Report

Quoted By: >>7383798

tried to clean up the fingers

Anonymous

Anonymous Fri 19 May 2023 14:23:54 No.7383797 Report

Quoted By:

i did a few different kinds of bakes and noticed that my loras tended to converge on what i wanted just after the 2nd cosine restart, even with 2x the steps, so i'm going to test a number of different step counts and see if i can find anything out

Anonymous

Anonymous Fri 19 May 2023 14:25:26 No.7383798 Report

Quoted By: >>7383801 >>7383803

>>7383782
>>7383795
Doing gods work.
I believe you can clean up those jpeg artifacts in photoshop, it has an AI filter for it.

Anonymous

Anonymous Fri 19 May 2023 14:29:09 No.7383801 Report

Quoted By: >>7383803 >>7383810

>>7383798
oof didnt see those before but they are pretty bad on the sky

Anonymous

Anonymous Fri 19 May 2023 14:30:54 No.7383803 Report

Quoted By: >>7383806

>>7383798
>>7383801
it's the part of aesthetic

Anonymous

View Same Google ImgOps iqdb SauceNAO 00010-4105288664.png, 1MiB, 1024x1536

Anonymous Fri 19 May 2023 14:32:18 No.7383805 Report

Quoted By: >>7383807

>>7383790

Anonymous

Anonymous Fri 19 May 2023 14:32:48 No.7383806 Report

Quoted By:

>>7383803
cel painting doesnt look like that homie

Anonymous

Anonymous Fri 19 May 2023 14:32:58 No.7383807 Report

Quoted By:

>>7383805
model?

Anonymous

Anonymous Fri 19 May 2023 14:36:29 No.7383810 Report

Quoted By:

>>7383801
the sky? damn, probably should've spent some more time in the Photoshop oven

Anonymous

View Same Google ImgOps iqdb SauceNAO xyz_grid-0002-830348164.jpg, 413KiB, 4000x609

Anonymous Fri 19 May 2023 14:39:30 No.7383812 Report

Quoted By: >>7383813 >>7383983

What could cause my style LoRA to mess up like this? Best guess is my dataset is not varied enough

Anonymous

Anonymous Fri 19 May 2023 14:41:14 No.7383813 Report

Quoted By: >>7383817

>>7383812
Too small of a dataset and bad tagging, likely.

Anonymous

Anonymous Fri 19 May 2023 14:44:15 No.7383817 Report

Quoted By: >>7383972 >>7384009

>>7383813
I ran wd14-swinv2-v2 at 0.5 and removed "anime coloring, official style, parody style" from the tags.

I hope it's the dataset, since this "prototype" was made on 188 images from only the first 16 episodes, so pulling more images would be easier than manually going over the tags

Anonymous

Anonymous Fri 19 May 2023 14:45:40 No.7383820 Report

Quoted By: >>7383825 >>7383827

So was that neutral prompt cfg update an actual thing? I think I saw something about further updates to it?

Anonymous

Anonymous Fri 19 May 2023 14:49:26 No.7383825 Report

Quoted By: >>7383830

>>7383820
CFG rescale will be superseded by https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/10555 soon (it's the same thing)
The perp-neg cfg algorithm the extension implements will still be applicable

Anonymous

Anonymous Fri 19 May 2023 14:50:16 No.7383827 Report

Quoted By: >>7383830

>>7383820
the math is currently not applied correctly i believe

Anonymous

Anonymous Fri 19 May 2023 14:51:40 No.7383830 Report

Quoted By: >>7383836

>>7383825
>>7383827
So just update and forget about it?

Anonymous

Anonymous Fri 19 May 2023 14:51:58 No.7383832 Report

Quoted By:

>reusing best gens for dataset to retrain character lora
>drops them into autotagger
>it recognises character right on all 16 gens
man

Anonymous

Anonymous Fri 19 May 2023 14:53:33 No.7383836 Report

Quoted By:

>>7383830
itll probably be fixed in a day or two

Anonymous

View Same Google ImgOps iqdb SauceNAO 00014.png, 4MiB, 1608x2412

Anonymous Fri 19 May 2023 14:57:42 No.7383843 Report

Quoted By: >>7383866 >>7384113

Anonymous

View Same Google ImgOps iqdb SauceNAO 00037-2427923868.png, 1MiB, 1024x1536

Anonymous Fri 19 May 2023 15:06:55 No.7383864 Report

Quoted By:

>>7383721

Anonymous

Anonymous Fri 19 May 2023 15:07:11 No.7383866 Report

Quoted By: >>7383877 >>7383880

>>7383843
I'd like a catbox, please

Anonymous

View Same Google ImgOps iqdb SauceNAO tmpwngdkge_.png, 1MiB, 1200x1800

Anonymous Fri 19 May 2023 15:11:04 No.7383873 Report

Quoted By:

Anonymous

Anonymous Fri 19 May 2023 15:12:30 No.7383877 Report

Quoted By: >>7384039

>>7383866
https://files.catbox.moe/u1qcau.png
inpainting before upscaling is my best friend

Anonymous

View Same Google ImgOps iqdb SauceNAO 00479-1221738016.png, 2MiB, 1152x1536

Anonymous Fri 19 May 2023 15:13:24 No.7383879 Report

Quoted By: >>7383886 >>7383946 >>7384264

Trying to do multiple characters without having to do a shitload of inpainting.

txt2img: https://files.catbox.moe/swdwq0.png

Anonymous

View Same Google ImgOps iqdb SauceNAO 00187-1106375878.png, 739KiB, 768x768

Anonymous Fri 19 May 2023 15:13:29 No.7383880 Report

Quoted By: >>7383883

>>7383866
not the same person but I found a fun keyword for making characters

Anonymous

View Same Google ImgOps iqdb SauceNAO 00206-1821952248.png, 767KiB, 768x768

Anonymous Fri 19 May 2023 15:17:56 No.7383883 Report

Quoted By:

>>7383880

Anonymous

Anonymous Fri 19 May 2023 15:18:03 No.7383886 Report

Quoted By: >>7383895

>>7383879
are you using regional prompter? it's pretty good. the tiled diffusion one doesn't mix them well enough in my experience

Anonymous

Anonymous Fri 19 May 2023 15:21:25 No.7383891 Report

Quoted By: >>7383898 >>7383903

Have AIs got good for generating double penetrations?

Anonymous

Anonymous Fri 19 May 2023 15:24:24 No.7383895 Report

Quoted By: >>7384011

>>7383886
No tricks, nothing extra. I'd only do regional prompting if I was trying to get two specific characters to keep their LoRAs from blending together. I've had success with that before, but it was for pokemon, which is anthro enough that I wouldn't be able to post anywhere but trash (and I'm not going to trash).

In this case, I want the characters to wear similar outfit. I really like showgirl skirts, but most checkpoints don't recognize it. Recently discovered "pussy peek" in positive and "clothes lift" in negative can get what I'm looking for somewhat consistently.

Only thing I'm using is tiled VAE so I have enough VRAM to upscale 2x and the add detail LoRA because I think it helps give better results.

Eventually I'm going to look into prompting for specific artists, but I don't look at enough hentai to recognize specific artists or know who makes good stuff (other than Sahara Wataru, I like his Naruto hentai).

Anonymous

Anonymous Fri 19 May 2023 15:27:37 No.7383898 Report

Quoted By: >>7383978

>>7383891
>double penetrations?
That's gay

Anonymous

Anonymous Fri 19 May 2023 15:29:58 No.7383903 Report

Quoted By:

>>7383891
you can controlnet it but in general ai doesn't do butt stuff veyr well

Anonymous

Anonymous Fri 19 May 2023 15:30:37 No.7383905 Report

Quoted By:

whats the current model everyone is using?

Just curious?

https://smartpolls.co.uk/p/31337

Anonymous

Anonymous Fri 19 May 2023 15:33:35 No.7383913 Report

Quoted By:

>>7383777
What style lora is this?

Anonymous

Anonymous Fri 19 May 2023 15:38:47 No.7383920 Report

Quoted By: >>7383926 >>7383929

>>7383536
what did you prompt for that side view?

Anonymous

Anonymous Fri 19 May 2023 15:41:20 No.7383926 Report

Quoted By:

>>7383920
its literally (from side)

Anonymous

Anonymous Fri 19 May 2023 15:45:05 No.7383929 Report

Quoted By:

>>7383920
there's been nothing aside of "reverse upright straddle" in terms of the view, so just rng or lora/model luck ( so still rng )

Anonymous

View Same Google ImgOps iqdb SauceNAO 00065-2345711987.png, 3MiB, 1216x1536

Anonymous Fri 19 May 2023 15:57:50 No.7383946 Report

Quoted By: >>7383966 >>7383993

>>7383879
How do people deal with same face?

Anonymous

Anonymous Fri 19 May 2023 16:13:27 No.7383966 Report

Quoted By: >>7383969

>>7383946
its pretty normal for anime so i dont think people care very much, you can just inpaint too

Anonymous

Anonymous Fri 19 May 2023 16:16:00 No.7383967 Report

Quoted By:

>>7383777
the only good poster itt

Anonymous

Anonymous Fri 19 May 2023 16:17:52 No.7383969 Report

Quoted By: >>7383979 >>7383993

>>7383966
>its pretty normal for anime to have same face
tsurime, tareme, jitome, sanpaku and many more eye types with their own tags on boorus: incomprehensible

Anonymous

Anonymous Fri 19 May 2023 16:19:41 No.7383970 Report

Quoted By:

I got have an AMD gpu cause better value in vidya. Does SD actually work well on amd? 6700xt.

Anonymous

Anonymous Fri 19 May 2023 16:20:18 No.7383972 Report

Quoted By: >>7383974

>>7383817
>0.5
anon...

Anonymous

View Same Google ImgOps iqdb SauceNAO 4032742312.png, 4MiB, 1536x1920

Anonymous Fri 19 May 2023 16:20:23 No.7383973 Report

Quoted By: >>7384013

Anonymous

Anonymous Fri 19 May 2023 16:20:52 No.7383974 Report

Quoted By: >>7383982 >>7383984 >>7384026

>>7383972
Is that not the recommended value for style?

Anonymous

Anonymous Fri 19 May 2023 16:22:45 No.7383978 Report

Quoted By:

>>7383898
based

Anonymous

Anonymous Fri 19 May 2023 16:22:57 No.7383979 Report

Quoted By:

>>7383969
wonderful strawman but i didnt say that

Anonymous

Anonymous Fri 19 May 2023 16:23:46 No.7383982 Report

Quoted By: >>7383995

>>7383974
Recommended by whom?
And hell nah.

Anonymous

Anonymous Fri 19 May 2023 16:23:46 No.7383983 Report

Quoted By:

>>7383812
looks like insane learning rates/repeats causing it to burn to a crisp?

Anonymous

Anonymous Fri 19 May 2023 16:24:05 No.7383984 Report

Quoted By: >>7383991

>>7383974
nta but id play around with it a bit, i think training suffers with too high (or too low) confidence. the default seems good, some other anon recommended 0.15 to me but it's too innacurate

Anonymous

View Same Google ImgOps iqdb SauceNAO Unou-12.png, 2MiB, 1024x1536

Anonymous Fri 19 May 2023 16:25:13 No.7383987 Report

Quoted By:

>train
>fullbody shots only reference one specific fullbody image in the dataset for some reason
weird

Anonymous

Anonymous Fri 19 May 2023 16:25:22 No.7383988 Report

Quoted By: >>7383999 >>7384022 >>7384027

anyone have recs on how to use neg prompts w inpainting? none of the guides ive seen mention it

Anonymous

Anonymous Fri 19 May 2023 16:26:18 No.7383991 Report

Quoted By: >>7383994

>>7383984
>too inaccurate
almost like you're supposed to go through these manually, dataset manager goes hard

Anonymous

View Same Google ImgOps iqdb SauceNAO 01282-639036120.png, 1MiB, 1536x1024

Anonymous Fri 19 May 2023 16:27:35 No.7383993 Report

Quoted By: >>7384007 >>7384012

>>7383969
If you use the Dynamic Prompts extension, you could put eye shapes into a wildcard file, but when you're only prompting "2girls", I don't know how effective that would be at giving them two different eye shapes. The problem of identical hairstyles is the one that really sticks out though. /trash/ likes to complain that NAI-based models always generate "M-shaped hair" (the official Danbooru tag is "hair between eyes"), which is really obvious in >>7383946, and one way you can avoid that is by having a wildcard file for bangs. Mine looks like this:

asymmetrical bangs
bangs pinned back
blunt bangs
crossed bangs
diagonal bangs
dyed bangs
hair over eyes
hair over one eye
long bangs
parted bangs
curtained hair
ribbon bangs
short bangs
swept bangs
hair intakes
sidelocks
asymmetrical sidelocks
low tied sidelocks
sidelocks tied back
single sidelock
widow's peak

Anonymous

Anonymous Fri 19 May 2023 16:27:52 No.7383994 Report

Quoted By: >>7384000

>>7383991
or you could select a slightly higher confidence and not have to remove a bunch of useless garbage like it tagging 15 random characters from random IPs

Anonymous

Anonymous Fri 19 May 2023 16:28:04 No.7383995 Report

Quoted By: >>7384005

>>7383982
Some dataset maker colab I've used ages ago recommended 0.3 for characters and 0.5 for styles. What do you recommend for styles instead?

Anonymous

View Same Google ImgOps iqdb SauceNAO pixiv.png, 35KiB, 1017x333

Anonymous Fri 19 May 2023 16:28:39 No.7383997 Report

Quoted By:

take the pixivpill today

Anonymous

Anonymous Fri 19 May 2023 16:29:21 No.7383999 Report

Quoted By: >>7384017

>>7383988
Try with (worst quality, low quality:1.4) and without and see what works better for you

Anonymous

Anonymous Fri 19 May 2023 16:29:28 No.7384000 Report

Quoted By: >>7384003

>>7383994
yeah but then you might have to manually tag what higher confidence missed

Anonymous

Anonymous Fri 19 May 2023 16:32:46 No.7384002 Report

Quoted By:

>>7383773
I'll try thanks

Anonymous

Anonymous Fri 19 May 2023 16:33:08 No.7384003 Report

Quoted By: >>7384006

>>7384000
training is essentially just telling the ai that some general tag means some specific collection of shapes and colors anyway, so you'd always want to do some research beforehand about which tags you're training on - except on a style, i'm always adding manual tags then appending autotagging. i don't have to worry about autotagging missing essential training elements because i never trust it to do that in the first place.

Anonymous

Anonymous Fri 19 May 2023 16:33:45 No.7384005 Report

Quoted By:

>>7383995
I use 0.15 for both.

Anonymous

Anonymous Fri 19 May 2023 16:35:18 No.7384006 Report

Quoted By: >>7384014 >>7384018

>>7384003
are you saying some things should not be tagged?

Anonymous

View Same Google ImgOps iqdb SauceNAO 00484-3636447770.png, 3MiB, 1216x1536

Anonymous Fri 19 May 2023 16:35:31 No.7384007 Report

Quoted By:

>>7383993
I still haven't messed around with wildcards, I'll look into it. I have noticed the M hair.

Anonymous

Anonymous Fri 19 May 2023 16:35:52 No.7384009 Report

Quoted By:

>>7383817
It depends on the dataset, but I'd always go lower than 0.5, so it doesn't miss a ton of details. Throw a couple of single images into the tagger and look what it's picking up and what it's missing. I think if you're training with a token size of 150, you can also test it by sending the tags from tagger to text2img and look if you're way above or below it. After you ran your batch through tagger, throw your folder in the BooruDatasetTagManager and mass prune tags that serve no purpose

Anonymous

Anonymous Fri 19 May 2023 16:36:21 No.7384011 Report

Quoted By: >>7384021

>>7383895
it looks actually pretty good with tiled, what parameters and upscalers do you use?

Anonymous

Anonymous Fri 19 May 2023 16:36:24 No.7384012 Report

Quoted By: >>7384034

>>7383993
what about explicitely specifying characters with needed expressions, clothes, face etc. on certain steps through prompt editing feature? https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Features#prompt-editing

Anonymous

Anonymous Fri 19 May 2023 16:36:33 No.7384013 Report

Quoted By:

>>7383973
Gimme the catbox bro, that looks dope

Anonymous

Anonymous Fri 19 May 2023 16:37:15 No.7384014 Report

Quoted By: >>7384150

>>7384006
no i'm saying you should tag everything but make sure you know what you're tagging

Anonymous

Anonymous Fri 19 May 2023 16:38:34 No.7384017 Report

Quoted By:

>>7383999
ty ill mess around w that

Anonymous

Anonymous Fri 19 May 2023 16:38:41 No.7384018 Report

Quoted By: >>7384150

>>7384006
for example i did a bunch of tests on how to get the ai to draw pubic barcode tattoos, i trained it on barcode, barcode tattoo, tattoo, pubic tattoo, pubic barcode tattoo; the clearest ended up being pubic tattoo, because it's always in the right location and there's no need to specify what the shape is since my input images do that.

Anonymous

View Same Google ImgOps iqdb SauceNAO catbox_9dnn4l.png, 3MiB, 1216x1536

Anonymous Fri 19 May 2023 16:41:02 No.7384021 Report

Quoted By: >>7384040

>>7384011
Here's the txt2img before inpainting; only had to fix the faces. Right now using 4x-AnimeSharp, but I use Remacri a lot too, and they're pretty interchangeable imo.

For Tiled VAE, everything is default except decoder tile size:
>Encoder Tile Size: 960 (left at default)
>Decoder Tile Size: 96 (that's as high as I can go before out of memory, the extension recommends going as high as possible)
>Fast Encoder and Fast Decoder checked

Anonymous

Anonymous Fri 19 May 2023 16:41:33 No.7384022 Report

Quoted By:

>>7383988
Depends on the type of inpainting.

Anonymous

Anonymous Fri 19 May 2023 16:47:44 No.7384026 Report

Quoted By: >>7384028 >>7384030 >>7384041

>>7383974
no anon, the best style loras I've used were flexible all the way from 0.3 to 1.3, also optimal if you're using it with an concept or scenery lora as to not overcook

same goes for characters, if the character shows all it's features at 0.8-1 without overcooking or influencing the model, it's good, since it means you can leave it at 0.8 as to show the character features correctly while changing prompt or other concept loras as you want

Anonymous

Anonymous Fri 19 May 2023 16:49:04 No.7384027 Report

Quoted By:

>>7383988
seconded

sometimes they say leave it blank, other times they say to put the whole prompt including seed

Anonymous

Anonymous Fri 19 May 2023 16:49:27 No.7384028 Report

Quoted By: >>7384044

>>7384026
Are you talking about weight in the prompt or WDTagger treshhold?

Anonymous

Anonymous Fri 19 May 2023 16:49:30 No.7384029 Report

Quoted By: >>7384043

What's the benefit of using an inpaint model, for inpainting?

Anonymous

Anonymous Fri 19 May 2023 16:49:50 No.7384030 Report

Quoted By: >>7384044

>>7384026
based dyslexic anon

Anonymous

View Same Google ImgOps iqdb SauceNAO 00919-853524993.png, 2MiB, 1536x1024

Anonymous Fri 19 May 2023 16:52:56 No.7384034 Report

Quoted By:

>>7384012
I'm not sure if that would work on two characters. The only reliable ways of making two different characters are with extensions like Regional Prompter or Multidiffusion. (That's not what I did here; I just prompted "3girls" and it generated four.)

Anonymous

Anonymous Fri 19 May 2023 17:01:07 No.7384039 Report

Quoted By:

>>7383877
Thank you
I'll try to keep up with the inpainting too

Anonymous

Anonymous Fri 19 May 2023 17:01:22 No.7384040 Report

Quoted By: >>7384057

>>7384021
thanks bud,

something else if you don't mind, what are your settings on the img2img/inpaint tab? desu a screencap will be enough, I'll try yours and compare with the settings I usually use

Anonymous

Anonymous Fri 19 May 2023 17:01:36 No.7384041 Report

Quoted By:

>>7384026
we're talking about tagging confidence, not lora weight

Anonymous

Anonymous Fri 19 May 2023 17:02:06 No.7384043 Report

Quoted By:

>>7384029
nothing you can just use a lora instead

Anonymous

Anonymous Fri 19 May 2023 17:04:15 No.7384044 Report

Quoted By:

>>7384028
>>7384030
LOL my bad, I just read "0.5" and thought he was asking about lora

Anonymous

View Same Google ImgOps iqdb SauceNAO file.png, 2KiB, 116x149

Anonymous Fri 19 May 2023 17:05:00 No.7384045 Report

Quoted By: >>7384048 >>7384049 >>7384051

what makes this happen every once in a while?
I generate a batch of 200 images and there's always like 12 of these

Anonymous

Anonymous Fri 19 May 2023 17:06:51 No.7384048 Report

Quoted By: >>7384052 >>7384056

>>7384045
Either stop using the NAI VAE or put --no-half-vae in your command line arguments, which would slow generation down slightly.

Anonymous

Anonymous Fri 19 May 2023 17:07:03 No.7384049 Report

Quoted By: >>7384052 >>7384056

>>7384045
stop using the shit nai vae, it's bugged and does this

Anonymous

Anonymous Fri 19 May 2023 17:08:56 No.7384051 Report

Quoted By: >>7384052

>>7384045
shit vae

Anonymous

Anonymous Fri 19 May 2023 17:09:59 No.7384052 Report

Quoted By: >>7384060

>>7384049
>>7384048
>>7384051
Oh, I see. What vae is actually good then because this has happened to me with a few that I've used from civitai so now I'm thinking they're probably the same shit with different names and I couldn't tell

Anonymous

Anonymous Fri 19 May 2023 17:12:18 No.7384056 Report

Quoted By: >>7384154

>>7384048
>>7384049
i like how nai vae looks on certain models, is there a similar looking vae that's not bugged? i don't want to use the command if it'll slow down generations especially since i tend to switch between vaes

tried clearvae and vae-ft-mse-840000-ema-pruned, they give errors too

Anonymous

View Same Google ImgOps iqdb SauceNAO Screenshot from 2023-05-19 13-07 (...).png, 95KiB, 944x860

Anonymous Fri 19 May 2023 17:12:20 No.7384057 Report

Quoted By: >>7385082

>>7384040
Is this what you want? The settings are basically the same as the original image, but that's because I was only fixing faces. If I was doing something more involved I only really change the denoising strength.

For the actual drawing of the inpaint mask, what I find to be helpful is look for areas where you won't mind inpainting lines to show up. For the last pic I posted, I drew the edge around the neck pieces. Then in the background, you can see there's support beams in the skylight, I tried to match the edge of the inpaint mask with those beams. This helps prevent discontinuities, or if they're unavoidable, put them in places where they don't matter.

Anonymous

Anonymous Fri 19 May 2023 17:13:23 No.7384060 Report

Quoted By:

>>7384052
sd 1.5 8400k or kl-f8-anime2

Anonymous

Anonymous Fri 19 May 2023 17:15:59 No.7384064 Report

Quoted By: >>7384065

Wait, I just realized that a lower treshhold means it recognized more tags instead of less...

Anonymous

Anonymous Fri 19 May 2023 17:17:21 No.7384065 Report

Quoted By:

>>7384064
Mhm. Common mistake. Did it myself.

Anonymous

Anonymous Fri 19 May 2023 17:21:00 No.7384070 Report

Quoted By: >>7384154

does anyone have the Lain Iwakura lora? The link in the op is dead

Anonymous

View Same Google ImgOps iqdb SauceNAO 00084-533221299-1girl, hair orna (...).png, 2MiB, 1280x1792

Anonymous Fri 19 May 2023 17:27:25 No.7384080 Report

Quoted By:

ugh
finally done

Anonymous

View Same Google ImgOps iqdb SauceNAO 00003-3092806925.png, 2MiB, 1536x1024

Anonymous Fri 19 May 2023 17:31:28 No.7384084 Report

Quoted By: >>7384088

Anonymous

View Same Google ImgOps iqdb SauceNAO 00161-2887271521-nsfw, scan, (tr (...).png, 2MiB, 1024x1536

Anonymous Fri 19 May 2023 17:33:54 No.7384086 Report

Quoted By:

Anonymous

Anonymous Fri 19 May 2023 17:34:22 No.7384088 Report

Quoted By: >>7384102

>>7384084
vae...
NOW!!!!!!!!

Anonymous

View Same Google ImgOps iqdb SauceNAO 00054-2631322901.png, 1MiB, 1280x1280

Anonymous Fri 19 May 2023 17:34:36 No.7384089 Report

Quoted By:

Anonymous

Anonymous Fri 19 May 2023 17:40:00 No.7384098 Report

Quoted By: >>7384106

how much colab time is colab pro? it says 100 compute units per month but nobody seems to know what the fuck that even means

Anonymous

View Same Google ImgOps iqdb SauceNAO 00059-1076488267.png, 1MiB, 1280x1280

Anonymous Fri 19 May 2023 17:40:53 No.7384101 Report

Quoted By: >>7384103

Anonymous

Anonymous Fri 19 May 2023 17:41:50 No.7384102 Report

Quoted By:

>>7384088
it has vae, not all light colors means forget vae

Anonymous

Anonymous Fri 19 May 2023 17:41:59 No.7384103 Report

Quoted By:

>>7384101
oh hey that's mine

Anonymous

Anonymous Fri 19 May 2023 17:44:54 No.7384106 Report

Quoted By: >>7384107

>>7384098
Running the SD webui consumes 2.05 compute units per hour. Last month, it was possible to keep using the webui after running out of compute units, but now Google may be getting serious about enforcing the quota. You may be able to keep running the notebook, or you could be shut down until the next billing cycle.

Anonymous

Anonymous Fri 19 May 2023 17:45:34 No.7384107 Report

Quoted By: >>7384119

>>7384106
Doesn't webui just not work anymore either way?

Anonymous

Anonymous Fri 19 May 2023 17:49:27 No.7384113 Report

Quoted By: >>7384122

>>7383843
Do you have a place where you post all your yordles anon? or just these threads?

Anonymous

Anonymous Fri 19 May 2023 18:00:52 No.7384119 Report

Quoted By: >>7384125

>>7384107
It's hard to predict if they'll stop you from running the notebook immediately after you exhaust your quota, or some random amount of time afterwards.

Anonymous

View Same Google ImgOps iqdb SauceNAO 00798-4140712794-(((outline))),1 (...).png, 813KiB, 768x1120

Anonymous Fri 19 May 2023 18:02:52 No.7384120 Report

Quoted By: >>7384121

Anonymous

Anonymous Fri 19 May 2023 18:03:31 No.7384121 Report

Quoted By:

>>7384120
That's a big rat.

Anonymous

View Same Google ImgOps iqdb SauceNAO 01149-3720701977.png, 1MiB, 1048x1568

Anonymous Fri 19 May 2023 18:03:34 No.7384122 Report

Quoted By: >>7384254

>>7384113
just here from time to time, i know there's another vexfucker in these threads though
i'm gonna rebake my poppy lora because last time i did was weeks ago and i tagged it like a moron so her armor is just glued to her skin

Anonymous

Anonymous Fri 19 May 2023 18:04:51 No.7384125 Report

Quoted By: >>7384136 >>7384137

>>7384119
I'm talking about the recent crackdown

Anonymous

View Same Google ImgOps iqdb SauceNAO 00802-3650679516-best quality, s (...).png, 820KiB, 768x1120

Anonymous Fri 19 May 2023 18:06:26 No.7384127 Report

Quoted By:

Anonymous

View Same Google ImgOps iqdb SauceNAO catbox_0gjqrl.jpg, 678KiB, 2667x4000

Anonymous Fri 19 May 2023 18:06:34 No.7384128 Report

Quoted By:

Anonymous

View Same Google ImgOps iqdb SauceNAO 00023-623518508.png, 2MiB, 1024x1536

Anonymous Fri 19 May 2023 18:10:32 No.7384131 Report

Quoted By:

Singing and lifting heavy metal

Anonymous

View Same Google ImgOps iqdb SauceNAO 00026-3680126536.png, 2MiB, 1024x1536

Anonymous Fri 19 May 2023 18:11:10 No.7384132 Report

Quoted By:

Anonymous

Anonymous Fri 19 May 2023 18:11:48 No.7384136 Report

Quoted By:

>>7384125
Google will not allow free Colab users to run the SD webui. If you pay them $11 per month for Colab Pro, they'll allow you to run it for 48 hours per month. It's unknown how much longer you can get away with running it after the 48 hours are up.

Anonymous

Anonymous Fri 19 May 2023 18:11:49 No.7384137 Report

Quoted By:

>>7384125
that's only for free users

Anonymous

View Same Google ImgOps iqdb SauceNAO 00488-3329282656.png, 2MiB, 1216x1536

Anonymous Fri 19 May 2023 18:14:31 No.7384144 Report

Quoted By: >>7384163 >>7384166

Anonymous

View Same Google ImgOps iqdb SauceNAO 1678597094693524-3965223047-Nats (...).png, 967KiB, 1024x1024

Anonymous Fri 19 May 2023 18:17:57 No.7384147 Report

Quoted By:

Anyone still bothering to controlnet porn screencaps? Which model would be best for that nowadays?

Anonymous

Anonymous Fri 19 May 2023 18:20:33 No.7384150 Report

Quoted By: >>7384162

>>7384018
>>7384014
thats just tag pruning, you'll be doing that regardless of WD strength and imo a waste of time to do it manually, since in your case you'd just use dataset manager to mass convert all strays into the one you wanted, way faster than doing it manually

Anonymous

Anonymous Fri 19 May 2023 18:22:17 No.7384154 Report

Quoted By: >>7384204 >>7384235

>>7384056
nai vae was bugfixed iirc and was available somewhere, try huggingface

>>7384070
civitai probably?

Anonymous

Anonymous Fri 19 May 2023 18:27:35 No.7384162 Report

Quoted By: >>7384172 >>7384179

>>7384150
its not faster to delete 15 random ass characters and fix 50 screenshots where the ai thinks there's "male focus" when there isn't. and i cant delete the tag entirely because that's not always wrong

Anonymous

Anonymous Fri 19 May 2023 18:28:22 No.7384163 Report

Quoted By:

>>7384144
that's a huge pussy

Anonymous

View Same Google ImgOps iqdb SauceNAO 00024-480112492ed.png, 952KiB, 1024x1536

Anonymous Fri 19 May 2023 18:28:53 No.7384165 Report

Quoted By: >>7384406

Anonymous

View Same Google ImgOps iqdb SauceNAO 00490-4011112655.png, 3MiB, 1216x1536

Anonymous Fri 19 May 2023 18:29:04 No.7384166 Report

Quoted By:

>>7384144

Anonymous

Anonymous Fri 19 May 2023 18:31:10 No.7384172 Report

Quoted By:

>>7384162
>take the few images with male focus
>invert selection
>delete tag
easy.

Anonymous

Anonymous Fri 19 May 2023 18:34:46 No.7384179 Report

Quoted By:

>>7384162
FWIW in all my tests, lowering the treshholds did not add any unnecessary tags.
Kinda opposite since it will often be super confident about wrong tags but then not tag the basics like sex positions, so lower ones are actually better.

Anonymous

Anonymous Fri 19 May 2023 18:48:01 No.7384204 Report

Quoted By: >>7384262

>>7384154
thanks but i must be retarded as i can't find it

animefull-final-pruned.vae and orangemix.vae are the same thing, right?

Anonymous

View Same Google ImgOps iqdb SauceNAO 9996490989765707-1347252404-kill (...).png, 2MiB, 2200x1600

Anonymous Fri 19 May 2023 18:53:26 No.7384217 Report

Quoted By: >>7384348

>>7383736
hopefully you prefer creampie killjoy

Anonymous

Anonymous Fri 19 May 2023 19:04:54 No.7384235 Report

Quoted By:

>>7384154
>civitai probably?
i couldn't find one on there maybe the huggingface hypernetwork will do what I want it to

Anonymous

View Same Google ImgOps iqdb SauceNAO Coom 'n Gloom.jpg, 208KiB, 1024x1536

Anonymous Fri 19 May 2023 19:11:19 No.7384254 Report

Quoted By: >>7384389

>>7384122
I'm not a vex fan or a foot guy, I'm just an inpaint enjoyer.

Anonymous

Anonymous Fri 19 May 2023 19:14:07 No.7384262 Report

Quoted By: >>7384275 >>7384302

>>7384204
no, orangemix is something else I think, AOM had the same vae as NAI

Anonymous

View Same Google ImgOps iqdb SauceNAO 00005-3433718496.png, 1MiB, 1024x1536

Anonymous Fri 19 May 2023 19:14:07 No.7384263 Report

Quoted By: >>7384266

For my first attempt at inpainting ever I have created something eldritch, something truly foul.

Like the good doctor Frankenstein, I now live in fear of my own deeds.

Anonymous

View Same Google ImgOps iqdb SauceNAO 00087-787351821.png, 2MiB, 1536x1024

Anonymous Fri 19 May 2023 19:14:44 No.7384264 Report

Quoted By:

>>7383879
I use regional prompter and inpainting.

Anonymous

Anonymous Fri 19 May 2023 19:15:08 No.7384266 Report

Quoted By:

>>7384263
And yes, I did inpaint the watermark back in

Anonymous

Anonymous Fri 19 May 2023 19:18:22 No.7384275 Report

Quoted By: >>7384287 >>7384302

>>7384262
retard

Anonymous

View Same Google ImgOps iqdb SauceNAO file.png, 453KiB, 1551x626

Anonymous Fri 19 May 2023 19:20:06 No.7384277 Report

Quoted By: >>7384313

>inpainting
Wow it actually did what I wanted it to for once and improved with each iteration.
Working with loras that had actual proper datasets (usually moege) is such a pleasurable experience compared to the usual mess I work with.

Anonymous

Anonymous Fri 19 May 2023 19:22:17 No.7384287 Report

Quoted By:

>>7384275
yeah I meant another, I keep mixing them up, sue me.

Anonymous

View Same Google ImgOps iqdb SauceNAO 00088-2879465527.png, 2MiB, 1536x1024

Anonymous Fri 19 May 2023 19:22:49 No.7384288 Report

Quoted By: >>7384293

Anonymous

Anonymous Fri 19 May 2023 19:24:47 No.7384293 Report

Quoted By: >>7384308

>>7384288
catbox? Art style is really nice.

Anonymous

Anonymous Fri 19 May 2023 19:26:18 No.7384294 Report

Quoted By: >>7384303

GANs are long overdue for a comeback
https://vcai.mpi-inf.mpg.de/projects/DragGAN/

Anonymous

View Same Google ImgOps iqdb SauceNAO 00395-1477963817.png, 1MiB, 1024x1536

Anonymous Fri 19 May 2023 19:27:48 No.7384297 Report

Quoted By:

shiranui flare

Anonymous

Anonymous Fri 19 May 2023 19:29:30 No.7384302 Report

Quoted By: >>7384305 >>7384307

>>7384262
thanks

>>7384275
at least specify why

Anonymous

View Same Google ImgOps iqdb SauceNAO DragGAN.webm, 2MiB, 400x316

Anonymous Fri 19 May 2023 19:29:38 No.7384303 Report

Quoted By: >>7384306 >>7384307 >>7384312 >>7384319

>>7384294

Anonymous

Anonymous Fri 19 May 2023 19:30:50 No.7384305 Report

Quoted By: >>7384325

>>7384302
what do you think the OM in AOM stands for newfag-kun?

Anonymous

Anonymous Fri 19 May 2023 19:30:57 No.7384306 Report

Quoted By:

>>7384303
I didn't want to use up an image for non /h/ material but aight anon

Anonymous

Anonymous Fri 19 May 2023 19:31:13 No.7384307 Report

Quoted By: >>7384325

>>7384303
seems far from perfect but cool nonetheless

>>7384302
>why
cause I mixed up AOM with Another, duh.

Anonymous

Anonymous Fri 19 May 2023 19:31:21 No.7384308 Report

Quoted By:

>>7384293
https://files.catbox.moe/gq1tox.png
I used inpainting to get rid of the errors around the eyes and pussy.

Anonymous

View Same Google ImgOps iqdb SauceNAO catbox_zzcs5b.png, 539KiB, 720x920

Anonymous Fri 19 May 2023 19:31:23 No.7384309 Report

Quoted By: >>7384316 >>7384322 >>7384350

>>7382620
Catbox nao

Anonymous

Anonymous Fri 19 May 2023 19:32:34 No.7384312 Report

Quoted By: >>7384315

>>7384303
bro this has been a more or less a thing for an actual decade

Anonymous

Anonymous Fri 19 May 2023 19:33:03 No.7384313 Report

Quoted By: >>7384350

>>7384277
did you edit the extra finger or that was pure inpainting?

Anonymous

Anonymous Fri 19 May 2023 19:33:32 No.7384315 Report

Quoted By: >>7384320

>>7384312
ain't no way

Anonymous

Anonymous Fri 19 May 2023 19:34:05 No.7384316 Report

Quoted By:

>>7384309
Aw shit that's the one I left the fucked up hand in

Anonymous

Anonymous Fri 19 May 2023 19:34:16 No.7384319 Report

Quoted By:

>>7384303
cool

Anonymous

Anonymous Fri 19 May 2023 19:34:45 No.7384320 Report

Quoted By:

>>7384315
Not dragging like this, but modifying images via blackbox sliders to do the same.

Anonymous

Anonymous Fri 19 May 2023 19:35:06 No.7384322 Report

Quoted By:

>>7384309
much inpainting

https://files.catbox.moe/4gf7nr.png

Anonymous

View Same Google ImgOps iqdb SauceNAO 00509-4157599317.png, 1MiB, 1024x1536

Anonymous Fri 19 May 2023 19:35:45 No.7384323 Report

Quoted By: >>7384342

shiranui flare

Anonymous

Anonymous Fri 19 May 2023 19:37:21 No.7384325 Report

Quoted By:

>>7384305
>>7384307
fucking kek
i'm retarded

Anonymous

View Same Google ImgOps iqdb SauceNAO 00427-1995290217.png, 1MiB, 1024x1536

Anonymous Fri 19 May 2023 19:37:52 No.7384326 Report

Quoted By: >>7384342

Anonymous

View Same Google ImgOps iqdb SauceNAO 00474-42984648.png, 2MiB, 1024x1536

Anonymous Fri 19 May 2023 19:38:58 No.7384330 Report

Quoted By: >>7384342

Anonymous

View Same Google ImgOps iqdb SauceNAO 00410-3839283373.png, 1MiB, 1024x1536

Anonymous Fri 19 May 2023 19:39:47 No.7384333 Report

Quoted By: >>7384342

Anonymous

View Same Google ImgOps iqdb SauceNAO 00226-1962142905.png, 2MiB, 1024x1536

Anonymous Fri 19 May 2023 19:41:02 No.7384335 Report

Quoted By: >>7384342

Anonymous

Anonymous Fri 19 May 2023 19:46:40 No.7384342 Report

Quoted By: >>7384377

>>7384323
>>7384326
>>7384330
>>7384333
>>7384335
These pictures are really clean looking. Can you catbox one so I can see your settings?

Anonymous

Anonymous Fri 19 May 2023 19:50:32 No.7384348 Report

Quoted By:

>>7384217
Very nice as always killjoybro

Anonymous

View Same Google ImgOps iqdb SauceNAO 1676735921389045-3248776232-Saki (...).png, 1MiB, 896x1024

Anonymous Fri 19 May 2023 19:51:38 No.7384350 Report

Quoted By: >>7384475

>>7384313
pic rel was initial, I removed some flesh on the left leg and that stray arm + extra fingers with Llama, manually smudged the colors and rough black outlines, then inpainted, seems it automatically understood a finger's missing so I took that and magic wanded the existing one a bit upwards to make space for the new one and then it was just inpainting and re-coloring the cock/outlines over and over since it tried to make it a double handjob by adding more fingers.
Pretty cool how it tried to re-add the knuckle I removed aswell. Some official art loras really hit different for hands

>>7384309
cool retro seibah

Anonymous

View Same Google ImgOps iqdb SauceNAO 23125-754456576-Yugi,__Yugi, sho (...).png, 1MiB, 1024x1536

Anonymous Fri 19 May 2023 19:53:33 No.7384355 Report

Quoted By: >>7384357 >>7384362 >>7384363

Adding Hidamari Sketch (anons want horrible styles for some reason) and Shinchou Yuusha

https://mega.nz/folder/xDkWWRAZ#jrvHsIeim3Q1HyDUEHLjbw

Anonymous

Anonymous Fri 19 May 2023 19:54:19 No.7384357 Report

Quoted By:

>>7384355
>horrible styles
soulful styles*

Anonymous

Anonymous Fri 19 May 2023 19:58:10 No.7384362 Report

Quoted By:

>>7384355
Your MEGA organization style is horrible. :^)
(thank u)

Anonymous

Anonymous Fri 19 May 2023 19:58:44 No.7384363 Report

Quoted By:

>>7384355
are you making these as paid requests?

Anonymous

View Same Google ImgOps iqdb SauceNAO 00848-1630716807-(masterpiece),( (...).png, 536KiB, 656x960

Anonymous Fri 19 May 2023 20:03:49 No.7384370 Report

Quoted By:

Anonymous

View Same Google ImgOps iqdb SauceNAO 00095-2096378373.png, 2MiB, 1536x1024

Anonymous Fri 19 May 2023 20:07:02 No.7384374 Report

Quoted By: >>7384376

Revisiting an old prompt.

Anonymous

Anonymous Fri 19 May 2023 20:07:41 No.7384376 Report

Quoted By:

>>7384374
I M A G I N E

Anonymous

Anonymous Fri 19 May 2023 20:08:31 No.7384377 Report

Quoted By:

>>7384342
Close to the default on this with Yesmix 15 and R-ESRGAN 4X+ Anime6B
https://civitai.com/models/27503/shiranui-flare-hololive

Anonymous

View Same Google ImgOps iqdb SauceNAO 00002-237894787-1girl, goblin, s (...).png, 2MiB, 1280x1792

Anonymous Fri 19 May 2023 20:09:30 No.7384378 Report

Quoted By: >>7384380 >>7384482

everyone is so mean to me...

Anonymous

Anonymous Fri 19 May 2023 20:09:41 No.7384379 Report

Quoted By: >>7384402

>>7382540
do i have to download anything to do this or is there a website?

Anonymous

Anonymous Fri 19 May 2023 20:10:27 No.7384380 Report

Quoted By: >>7384381

>>7384378
its because you keep making ugly gens

have you tried something even vaguely close to a conventionally beautiful female human

Anonymous

Anonymous Fri 19 May 2023 20:11:04 No.7384381 Report

Quoted By: >>7384387

>>7384380
Nu uh.

Anonymous

View Same Google ImgOps iqdb SauceNAO 163272-1121221216, DPM++ 2M Karr (...).png, 1MiB, 1280x880

Anonymous Fri 19 May 2023 20:11:52 No.7384382 Report

Quoted By: >>7384385 >>7384393 >>7384401

"Well? It's not going to suck itself"

Anonymous

Anonymous Fri 19 May 2023 20:13:09 No.7384385 Report

Quoted By: >>7384396

>>7384382
I've gotten one cunnilingus pov in my life and it was the highlight of me gens.

Anonymous

Anonymous Fri 19 May 2023 20:14:11 No.7384387 Report

Quoted By: >>7384390

>>7384381
maybe if you made something that isn't a short fat midget people would like it

Anonymous

View Same Google ImgOps iqdb SauceNAO 00604-4012214757.png, 2MiB, 1024x1536

Anonymous Fri 19 May 2023 20:14:41 No.7384388 Report

Quoted By:

Anonymous

View Same Google ImgOps iqdb SauceNAO 01150-392611125.png, 2MiB, 1048x1568

Anonymous Fri 19 May 2023 20:14:47 No.7384389 Report

Quoted By:

>>7384254
i can respect that

Anonymous

Anonymous Fri 19 May 2023 20:14:56 No.7384390 Report

Quoted By:

>>7384387
Nuh uh.

Anonymous

View Same Google ImgOps iqdb SauceNAO 00601-418909891.png, 2MiB, 1024x1536

Anonymous Fri 19 May 2023 20:15:25 No.7384391 Report

Quoted By:

Anonymous

Anonymous Fri 19 May 2023 20:17:41 No.7384393 Report

Quoted By: >>7384484

>>7384382
well my mouth isnt gonna be full of piss by itself

Anonymous

Anonymous Fri 19 May 2023 20:18:03 No.7384396 Report

Quoted By:

>>7384385
red

Anonymous

Anonymous Fri 19 May 2023 20:22:41 No.7384401 Report

Quoted By: >>7384409

>>7384382
how?

Anonymous

Anonymous Fri 19 May 2023 20:23:42 No.7384402 Report

Quoted By:

>>7384379
/g/ has a better non-vandalized OP with a guide. You gotta download but theres shitty sites that offer this for free, just worse

Anonymous

View Same Google ImgOps iqdb SauceNAO 00012-1751728384-1girl, shortsta (...).png, 2MiB, 1288x1792

Anonymous Fri 19 May 2023 20:25:23 No.7384404 Report

Quoted By:

horny midgets in your area

Anonymous

Anonymous Fri 19 May 2023 20:28:42 No.7384406 Report

Quoted By: >>7384920

>>7384165
i like it, simple and cute
what prompt gets reimu in that pose?

Anonymous

View Same Google ImgOps iqdb SauceNAO 00888-3666836001-, def1,bare sho (...).png, 637KiB, 768x1152

Anonymous Fri 19 May 2023 20:29:56 No.7384408 Report

Quoted By:

Anonymous

Anonymous Fri 19 May 2023 20:31:07 No.7384409 Report

Quoted By: >>7384412

>>7384401
looks like POV facesitting lora on civitai, or good prompts

https://civitai.com/models/37343/pov-facesitting

Anonymous

Anonymous Fri 19 May 2023 20:33:15 No.7384412 Report

Quoted By:

>>7384409
3D shit, doesnt seem to be it.
I think hes using the detail lora too

Anonymous

View Same Google ImgOps iqdb SauceNAO 00003-3726898574.png, 1MiB, 1024x1536

Anonymous Fri 19 May 2023 20:45:18 No.7384431 Report

Quoted By:

Thanks for the advice guys, lora turned out pretty good

Anonymous

View Same Google ImgOps iqdb SauceNAO catbox_7b5kqr.png, 2MiB, 1536x1024

Anonymous Fri 19 May 2023 20:47:20 No.7384438 Report

Quoted By: >>7384439 >>7384458

>git good rolls on hands
>dagone hat turns into a half horn thing
guess shes got horns now..

Anonymous

Anonymous Fri 19 May 2023 20:47:45 No.7384439 Report

Quoted By: >>7384445

>>7384438
inpaint

Anonymous

Anonymous Fri 19 May 2023 20:50:46 No.7384445 Report

Quoted By: >>7384453 >>7384456 >>7384457

>>7384439
I spent like 20 mins on it, and nothing would work or blend in. Plus im kinda retarded.

Anonymous

Anonymous Fri 19 May 2023 20:52:20 No.7384447 Report

Quoted By: >>7384449

openpose with penis bone when?

Anonymous

Anonymous Fri 19 May 2023 20:53:17 No.7384449 Report

Quoted By:

>>7384447
when humans evolve further

Anonymous

Anonymous Fri 19 May 2023 20:57:07 No.7384453 Report

Quoted By: >>7384456

>>7384445
>Plus im kinda retarded
i mean yeah but inpaint is ez.

Anonymous

Anonymous Fri 19 May 2023 20:59:25 No.7384456 Report

Quoted By: >>7384463

>>7384445
>>7384453 me
>that catbox
Okay, it might not work if you're THAT retarded. Never mind.

Anonymous

Anonymous Fri 19 May 2023 21:00:21 No.7384457 Report

Quoted By: >>7384469 >>7385101

>>7384445
inpaint the area you want to change, adjust the denoise before rendering (higher results in more drastic change), use sample whole image if you dont want to fuck with the prompt
i honestly suggest not doing hires upon gens, instead get your 512x768 or whatever and inpaint to fix major items and then upres in img2img

Anonymous

View Same Google ImgOps iqdb SauceNAO 70 inpaints for a hat.webm, 3MiB, 1280x720

Anonymous Fri 19 May 2023 21:00:20 No.7384458 Report

Quoted By: >>7384464 >>7384471 >>7384475

>>7384438
Yup. sounds about right.

Anonymous

Anonymous Fri 19 May 2023 21:03:03 No.7384463 Report

Quoted By: >>7384466

>>7384456
>THAT retarded
Hop off cock, nigga. I just started this shit 2 days ago, never done media editing of any kind

Anonymous

Anonymous Fri 19 May 2023 21:03:36 No.7384464 Report

Quoted By: >>7384483

>>7384458
just copy and paste a hat, like from a 3d model, inpaint at low denoise, done
this gacha shit people do is not good

Anonymous

Anonymous Fri 19 May 2023 21:06:41 No.7384466 Report

Quoted By: >>7384476

>>7384463
Who said anything about media editing. Your prompts are a niggardly mess. Fix yourself before you break yourself.

Anonymous

Anonymous Fri 19 May 2023 21:07:42 No.7384469 Report

Quoted By:

>>7384457
Thanks for the tips G, i only res'd up when i accepted the fate of the pic

Anonymous

Anonymous Fri 19 May 2023 21:09:38 No.7384471 Report

Quoted By: >>7384483

>>7384458
Someone get this man a faux hat for corns sake

Anonymous

View Same Google ImgOps iqdb SauceNAO CharaStudio-2021-01-20-02-53-47- (...).png, 2MiB, 1920x1080

Anonymous Fri 19 May 2023 21:11:27 No.7384474 Report

Quoted By:

any good models/loras for rimjobs?

Anonymous

Anonymous Fri 19 May 2023 21:11:53 No.7384475 Report

Quoted By: >>7384478 >>7384618

>>7384350
nice, i'm assuming that was inpainted with whole area selected?
i feel like inpainting is also highly dependent on the model, i'm trying out inpainting with different checkpoints and it seems to be easier in some than others

>>7384458
based, i like it when the end result doesn't have any obvious traces of inpainting like with the most of your gens that i've seen

Anonymous

Anonymous Fri 19 May 2023 21:12:41 No.7384476 Report

Quoted By:

>>7384466
>chek'd
i didn't even know styles added to prompt lists. I have only about 8-10 written in the actual prompt area.

Anonymous

View Same Google ImgOps iqdb SauceNAO 00020-3968700698.png, 802KiB, 768x768

Anonymous Fri 19 May 2023 21:14:07 No.7384477 Report

Quoted By: >>7384533

First lora I baked for Uni didn't come out so good. Proportions are all wrong, outfit is off and got none of the colors, didn't get the cross pupils, hair ribbon turned into extra wings. Trimmed the data set from 90 to 60 removing most alternate outfits, upped repetitions from 5 to 10, and rebaking.

Anonymous

Anonymous Fri 19 May 2023 21:14:18 No.7384478 Report

Quoted By: >>7384556 >>7385097

>>7384475
>i feel like inpainting is also highly dependent on the model, i'm trying out inpainting with different checkpoints and it seems to be easier in some than others
I don't think that makes logical sense other than if the model sucks in the first place. It's the same genning, isn't it?

Anonymous

Anonymous Fri 19 May 2023 21:15:26 No.7384481 Report

Quoted By:

oh yeah i was supposed to check out adafactor

Anonymous

Anonymous Fri 19 May 2023 21:15:28 No.7384482 Report

Quoted By: >>7384485

>>7384378
not even the worst goblinposter

Anonymous

Anonymous Fri 19 May 2023 21:16:15 No.7384483 Report

Quoted By:

>>7384464
>>7384471
I actually don't gacha at all. I use photoshop to draw in details and such and img2img and inpainting to blend them in. I just did this for fun.

Anonymous

Anonymous Fri 19 May 2023 21:16:23 No.7384484 Report

Quoted By:

>>7384393
based

Anonymous

Anonymous Fri 19 May 2023 21:17:00 No.7384485 Report

Quoted By:

>>7384482
I guess that's somewhat of a compliment...

Anonymous

Anonymous Fri 19 May 2023 21:52:20 No.7384508 Report

Quoted By: >>7384509

I thought SAG was supposed to work without patching now but it doesn't? Is it only fixed on the dev branch?

Anonymous

Anonymous Fri 19 May 2023 21:53:47 No.7384509 Report

Quoted By: >>7384514

>>7384508
dev branch no longer requires the manual patch now

Anonymous

View Same Google ImgOps iqdb SauceNAO 00783-2456158913.png, 1MiB, 1024x1536

Anonymous Fri 19 May 2023 21:56:58 No.7384513 Report

Quoted By: >>7384524

Anonymous

Anonymous Fri 19 May 2023 21:57:01 No.7384514 Report

Quoted By: >>7384519

>>7384509
rip
wish it would be merged into master; the manual patch only works on that specific commit and that commit is shit

Anonymous

View Same Google ImgOps iqdb SauceNAO 00377-1438956795.png, 1MiB, 1024x1536

Anonymous Fri 19 May 2023 21:57:35 No.7384515 Report

Quoted By: >>7384524

Anonymous

View Same Google ImgOps iqdb SauceNAO 00366-1711415501.png, 1MiB, 1024x1536

Anonymous Fri 19 May 2023 21:58:39 No.7384517 Report

Quoted By: >>7384524

Anonymous

View Same Google ImgOps iqdb SauceNAO 00344-3204616902.png, 1MiB, 1024x1536

Anonymous Fri 19 May 2023 21:59:14 No.7384518 Report

Quoted By: >>7384524

Anonymous

Anonymous Fri 19 May 2023 21:59:42 No.7384519 Report

Quoted By: >>7384526

>>7384514
It will be merged soon enough. If you know what you're doing it's easy enough to just git checkout dev. Full list of upcoming stuff here: https://github.com/AUTOMATIC1111/stable-diffusion-webui/blob/dev/CHANGELOG.md
>that specific commit and that commit is shit
get ready for the (You)s you're about to get on this, lol.

Anonymous

View Same Google ImgOps iqdb SauceNAO 00321-3528406545.png, 2MiB, 1024x1536

Anonymous Fri 19 May 2023 21:59:49 No.7384520 Report

Quoted By:

Anonymous

View Same Google ImgOps iqdb SauceNAO 00313-3236037459.png, 2MiB, 1024x1536

Anonymous Fri 19 May 2023 22:00:24 No.7384521 Report

Quoted By: >>7384524

Anonymous

View Same Google ImgOps iqdb SauceNAO 00265-1428370667.png, 1MiB, 1024x1536

Anonymous Fri 19 May 2023 22:00:59 No.7384523 Report

Quoted By: >>7384525

Anonymous

Anonymous Fri 19 May 2023 22:00:58 No.7384524 Report

Quoted By:

>>7384521
>>7384518
>>7384517
>>7384515
>>7384513
my wife

Anonymous

Anonymous Fri 19 May 2023 22:03:03 No.7384525 Report

Quoted By:

>>7384523
So close to a panty peek

Anonymous

Anonymous Fri 19 May 2023 22:04:03 No.7384526 Report

Quoted By: >>7384531

>>7384519
Yeah I just didn't wanna bother with it in case it was broken somehow.
>allow choosing sampler and prompts for second pass of hires fix - hidden by default, enabled in settings
Finally.

Anonymous

Anonymous Fri 19 May 2023 22:06:02 No.7384529 Report

Quoted By: >>7384530 >>7384550

Is img2img:ing some photos into drawings, and then using said drawings to train a LoRA, a valid approach? I guess that training an anime model with photos would just create weirdness... or?

I want to gen unnaturally big, hard, beach ball shaped bolt-on bimbo tits, but prompts alone don't cut it. Finding hentai source material has proven harder than I thought but regular 3dpd, especially mature vintage, has plenty of them.

Anonymous

Anonymous Fri 19 May 2023 22:07:43 No.7384530 Report

Quoted By: >>7384540

>>7384529
if you generate them with your baseline model without using a negative prompt or vae it should be ok, but also why

Anonymous

Anonymous Fri 19 May 2023 22:08:14 No.7384531 Report

Quoted By: >>7384536

>>7384526
fwiw I use the dev branch exclusively now and have had no issues. to be fair though everyone was essentially subject to that prior and I also had no issues during that time either

Anonymous

View Same Google ImgOps iqdb SauceNAO 4033032826.png, 1MiB, 1000x1000

Anonymous Fri 19 May 2023 22:09:39 No.7384532 Report

Quoted By: >>7384537

Flare?

Anonymous

View Same Google ImgOps iqdb SauceNAO 00005-157780267.png, 1MiB, 1024x1024

Anonymous Fri 19 May 2023 22:10:51 No.7384533 Report

Quoted By: >>7384635 >>7384922

>>7384477
Huge improvement but still very sloppy. Will try to keep fine tuning.

Anonymous

Anonymous Fri 19 May 2023 22:11:12 No.7384534 Report

Quoted By: >>7384539

>NansException: A tensor with all NaNs was produced in VAE. This could be because there's not enough precision to represent the picture. Try adding --no-half-vae commandline argument to fix this. Use --disable-nan-check commandline argument to disable this check.
What is the best way to fix this? I'm running on a 3060 and don't have the VRAM to do arguments.

Anonymous

Anonymous Fri 19 May 2023 22:14:34 No.7384536 Report

Quoted By: >>7385792

>>7384531
>do git checkout dev
>won't even launch because of operand TypeError
thank

Anonymous

View Same Google ImgOps iqdb SauceNAO 00813-3756478052.png, 2MiB, 1024x1536

Anonymous Fri 19 May 2023 22:16:04 No.7384537 Report

Quoted By:

>>7384532
hell yeah

Anonymous

Anonymous Fri 19 May 2023 22:19:08 No.7384539 Report

Quoted By:

>>7384534
use a different vae

Anonymous

Anonymous Fri 19 May 2023 22:19:33 No.7384540 Report

Quoted By: >>7384560

>>7384530
If you mean why this approach, I can't find 2D source material. If you're questioning my fetish, no bully pls

Thanks for the tips about baseline models, no vae or negative prompts. I have a surface level understanding but I guess a VAE and negative prompts can fuck with the latent space in subtle ways which screws up your lora...?

Anonymous

View Same Google ImgOps iqdb SauceNAO 00144-3852828540.png, 2MiB, 1024x1536

Anonymous Fri 19 May 2023 22:21:55 No.7384544 Report

Quoted By: >>7384546 >>7384571

Anonymous

Anonymous Fri 19 May 2023 22:23:21 No.7384546 Report

Quoted By: >>7384554

>>7384544
You might want to add a negative prompt for watermark/username/pateron etc

Anonymous

Anonymous Fri 19 May 2023 22:25:55 No.7384547 Report

Quoted By: >>7384580

why does the booru dataset tag manager crash upon trying to load any folder, the webui dataset manager is fucking ass to use

Anonymous

View Same Google ImgOps iqdb SauceNAO 00153.png, 2MiB, 1200x1800

Anonymous Fri 19 May 2023 22:25:59 No.7384548 Report

Quoted By: >>7384793 >>7385487

>>7381516
And? Did it work?

Anonymous

Anonymous Fri 19 May 2023 22:26:35 No.7384550 Report

Quoted By:

>>7384529
Not really?

Anonymous

Anonymous Fri 19 May 2023 22:29:08 No.7384554 Report

Quoted By:

>>7384546
sigh I missed this one. I already have 4 tokens specifically devoted to that. I'll add patreon and username too. Thanks.

Anonymous

Anonymous Fri 19 May 2023 22:29:59 No.7384556 Report

Quoted By:

>>7384478
from my experience it was easier to inpaint certain things with simpler artstyles, especially hands and fingers if they're in weird positions. i have tried some of the mixes with a little more realism, not full blown 2.5d, and they can make inpainting a pain in the ass. inpaint a body part on too low of a resolution and it will keep fucking them but going for higher resolutions will add too many realistic details and can make it look off even with realistic etc in negatives, drawn styles or models might not have that sort of issue but the more details there are the more you potentially might need to fix and i guess that gives more room for ai to fuck up. but all of that depends on the level of autism you're willing to throw into gens

i suppose "model sucks" is a good way to put it, some models seemed to be way more responsive to my prompts when inpainting, both whole picture and only masked area, but it could be just some incredible gacha

i should probably spend more time with inpainting variants

Anonymous

Anonymous Fri 19 May 2023 22:31:57 No.7384560 Report

Quoted By:

>>7384540
>Thanks for the tips about baseline models, no vae or negative prompts. I have a surface level understanding but I guess a VAE and negative prompts can fuck with the latent space in subtle ways which screws up your lora...?
it's because you only want it to learn the content of the training images, not the style. if the style matches the baseline model's then it can't really learn the style. this is important since all of your gens are going to be from a single source.

Anonymous

View Same Google ImgOps iqdb SauceNAO 03456-199752317.png, 1MiB, 864x1152

Anonymous Fri 19 May 2023 22:38:35 No.7384567 Report

Quoted By:

>>7382753
the fishine lora is really fucking good at resting bitch face, I should use it more often even at low strength with other loras

Anonymous

Anonymous Fri 19 May 2023 22:42:14 No.7384571 Report

Quoted By: >>7384623

>>7384544
Waifus sure are magnets for birds

Anonymous

View Same Google ImgOps iqdb SauceNAO 01151-1883513377.png, 2MiB, 1072x1608

Anonymous Fri 19 May 2023 22:50:20 No.7384575 Report

Quoted By:

im terminal

Anonymous

View Same Google ImgOps iqdb SauceNAO vexxo.jpg, 206KiB, 1024x1536

Anonymous Fri 19 May 2023 22:55:56 No.7384579 Report

Quoted By: >>7384834

Anonymous

Anonymous Fri 19 May 2023 22:57:22 No.7384580 Report

Quoted By:

>>7384547
apparently gradio updates fuck up the speed so people use an older auto1111 and dataset version, this is a quick install to the latest version before it broke https://github.com/toshiaki1729/dataset-tag-editor-standalone

Anonymous

Anonymous Fri 19 May 2023 23:01:54 No.7384583 Report

Quoted By:

>>7383444
Now this is actually based

Anonymous

View Same Google ImgOps iqdb SauceNAO 00209-2098712367-nsfw, scan, (tr (...).png, 2MiB, 1024x1536

Anonymous Fri 19 May 2023 23:04:06 No.7384585 Report

Quoted By: >>7384593 >>7384598

Anonymous

Anonymous Fri 19 May 2023 23:10:49 No.7384593 Report

Quoted By:

>>7384585
cute clit

Anonymous

Anonymous Fri 19 May 2023 23:19:44 No.7384598 Report

Quoted By:

>>7384585
get pregnant!

Anonymous

View Same Google ImgOps iqdb SauceNAO 00915-3215035160.png, 2MiB, 1024x1536

Anonymous Fri 19 May 2023 23:22:16 No.7384602 Report

Quoted By: >>7384607

Anonymous

Anonymous Fri 19 May 2023 23:23:59 No.7384604 Report

Quoted By: >>7384605

decided to dowload and try a couple of vaes for myself but i'll just leave this here, didn't use xformers and no upscaling
the shirt is different on aom3 because it decided to give me an uoh surprise, had to manually make it more presentable

non compressed png here
https://files.catbox.moe/etjzfy.png

metadata
https://files.catbox.moe/raxt10.png

Anonymous

View Same Google ImgOps iqdb SauceNAO testingvae.jpg, 4MiB, 6912x7426

Anonymous Fri 19 May 2023 23:24:35 No.7384605 Report

Quoted By:

>>7384604
retard

Anonymous

View Same Google ImgOps iqdb SauceNAO 00918-4198948680.png, 2MiB, 1024x1536

Anonymous Fri 19 May 2023 23:25:19 No.7384607 Report

Quoted By: >>7384613

>>7384602
alternate

Anonymous

View Same Google ImgOps iqdb SauceNAO file.png, 3MiB, 2552x856

Anonymous Fri 19 May 2023 23:27:07 No.7384609 Report

Quoted By: >>7384614

Is there a way to have img2img dont turn down the colors in an image, after a few iterations?

Anonymous

Anonymous Fri 19 May 2023 23:27:57 No.7384610 Report

Quoted By:

as usual, literally no reason to use anything besides 840000

Anonymous

View Same Google ImgOps iqdb SauceNAO 00160.png, 2MiB, 2016x1344

Anonymous Fri 19 May 2023 23:30:46 No.7384613 Report

Quoted By: >>7384616 >>7384617

>>7384607
nice clean pics there anon, catbox please?

Anonymous

Anonymous Fri 19 May 2023 23:31:26 No.7384614 Report

Quoted By:

>>7384609
no

Anonymous

Anonymous Fri 19 May 2023 23:35:27 No.7384616 Report

Quoted By: >>7384617

>>7384613
https://files.catbox.moe/zn52ew.png
https://files.catbox.moe/zn52ew.png

Anonymous

Anonymous Fri 19 May 2023 23:36:28 No.7384617 Report

Quoted By:

>>7384613
>>7384616
Oops, messed that up
https://files.catbox.moe/ajhyr0.png

Anonymous

View Same Google ImgOps iqdb SauceNAO file.png, 338KiB, 897x891

Anonymous Fri 19 May 2023 23:36:53 No.7384618 Report

Quoted By: >>7384631

how important/damaging are stray tags? this list goes down to 2-3x as long and im sure as hell not checking every tag, it should be fine if one picture's mistagged, no?

>>7384475
yep, was inpainting whole img
>also highly dependent on the model
it is yeah, but it varies on case by case basis, one model wont always be better than the other

Anonymous

View Same Google ImgOps iqdb SauceNAO 00210-729020269.png, 2MiB, 1024x1536

Anonymous Fri 19 May 2023 23:44:21 No.7384623 Report

Quoted By:

>>7384571
I just don't know why these birds are so careless when eating their yogurt.

Anonymous

Anonymous Fri 19 May 2023 23:55:00 No.7384631 Report

Quoted By: >>7384656

>>7384618
i dont think theres anything wrong as long as the tagger wasnt wrong

Anonymous

View Same Google ImgOps iqdb SauceNAO 00955-574384155.png, 1MiB, 1024x1536

Anonymous Fri 19 May 2023 23:58:59 No.7384633 Report

Quoted By:

Anonymous

Anonymous Sat 20 May 2023 00:00:29 No.7384634 Report

Quoted By: >>7384636 >>7384657

Will the lora baker crash if I go over vram or something? Can I still use my pc without fucking it up?

Anonymous

Anonymous Sat 20 May 2023 00:01:18 No.7384635 Report

Quoted By:

>>7384533
based alexianon

Anonymous

Anonymous Sat 20 May 2023 00:01:35 No.7384636 Report

Quoted By: >>7384640

>>7384634
saturating VRAM will start to page it in and out of other backing storage like RAM which will massively throttle performance, potentially deadlocking or crashing.
would not recommmend

Anonymous

View Same Google ImgOps iqdb SauceNAO 00964-3713032930.png, 1MiB, 1024x1536

Anonymous Sat 20 May 2023 00:07:05 No.7384637 Report

Quoted By:

Anonymous

Anonymous Sat 20 May 2023 00:11:30 No.7384640 Report

Quoted By: >>7384657

>>7384636
what's the most efficient way to bake at 768 so I get some leeway?

Anonymous

View Same Google ImgOps iqdb SauceNAO 00972-1940780453.png, 1MiB, 1024x1536

Anonymous Sat 20 May 2023 00:18:36 No.7384647 Report

Quoted By:

Anonymous

View Same Google ImgOps iqdb SauceNAO 00246-3796074006.png, 2MiB, 1024x1536

Anonymous Sat 20 May 2023 00:24:40 No.7384651 Report

Quoted By: >>7384680

Anonymous

Anonymous Sat 20 May 2023 00:27:57 No.7384656 Report

Quoted By: >>7384664 >>7384675 >>7384682 >>7385289

>>7384631
im specifically asking about wrong tags, stuff like jacket tagged as coat or some extra tags such as tea cup tagged as tea pot etc
then again I doubt anyone actuallyran tests with some tag and without it

Anonymous

Anonymous Sat 20 May 2023 00:28:57 No.7384657 Report

Quoted By: >>7384713

>>7384634
it will usually just tell you that you dont have enough vram instantly

>>7384640
how much does your gpu have

Anonymous

View Same Google ImgOps iqdb SauceNAO 00982-2629480093.png, 2MiB, 1024x1536

Anonymous Sat 20 May 2023 00:29:57 No.7384660 Report

Quoted By:

Anonymous

Anonymous Sat 20 May 2023 00:31:16 No.7384661 Report

Quoted By: >>7384663 >>7384692 >>7384698 >>7384730

vtai is taking pity on us tonight and saving this hellhole, arigatou.
haachama when?

Anonymous

Anonymous Sat 20 May 2023 00:32:06 No.7384663 Report

Quoted By: >>7384666

>>7384661
>vtai
?

Anonymous

Anonymous Sat 20 May 2023 00:32:34 No.7384664 Report

Quoted By: >>7384676

>>7384656
well now that you mention it i would be pretty easy to test trying to poison your lora with bad tagging

Anonymous

Anonymous Sat 20 May 2023 00:32:52 No.7384666 Report

Quoted By: >>7384669

>>7384663
the vtuber AI thread on the vtuber board

Anonymous

Anonymous Sat 20 May 2023 00:33:55 No.7384669 Report

Quoted By:

>>7384666
gross

Anonymous

Anonymous Sat 20 May 2023 00:35:14 No.7384672 Report

Quoted By: >>7384722

using to colab lora training script from the op. getting this error

> RuntimeError: Detected that PyTorch and torchvision were compiled with different CUDA versions. PyTorch has CUDA Version=11.7 and torchvision has CUDA Version=11.8. Please reinstall the torchvision that matches your PyTorch install.

Anyone have a fix for this?

Anonymous

Anonymous Sat 20 May 2023 00:36:24 No.7384675 Report

Quoted By: >>7384677

>>7384656
>then again I doubt anyone actuallyran tests with some tag and without it
are you aware that this stuff has been around since december

Anonymous

Anonymous Sat 20 May 2023 00:36:42 No.7384676 Report

Quoted By:

>>7384664
it would be extremely tedious to test properly as you'd need to try stuff like taking 10 images and not tagging a couch in one of them, then doing several tests to see if this causes the couch to appear fucking up backgrounds. If yes you would raise images to 30 and repeat, then 60, then 120 - by which point it should be enough to tell what effect a single picture will have.
Oh and you'd have to do this with multiple image sets just to make sure this isnt limited to a single image.
And dont forget you would have to retrain multiple times on different seeds for each of these test cases.
Then you would have to test whether batch sizes affect this aswell.
And then you could move on to having more than one fucked up tag.
Very easy. Granted you could probably skip a few of these steps once you confirm that there's no point to them, but just weeding the possibilities out would take insane amounts of time

Anonymous

Anonymous Sat 20 May 2023 00:39:41 No.7384677 Report

Quoted By: >>7384686

>>7384675
Yeah, and since I've been here all this time, I would most likely have known if someone did test this properly. You only get tidbits like some guy saying a single image fucked up his entire lora, someone mentioning a loose tag doing it, but from my experience its been extremely inconsistent - sometimes you can have untagged shit that wont appear, sometimes you can tag it and it will still appear everywhere.
Its making me question what is the actually correct tagging practice and if it even matters much - i.e bad tagging not existing, only bad images.

Anonymous

View Same Google ImgOps iqdb SauceNAO cooom.png, 1MiB, 1024x1024

Anonymous Sat 20 May 2023 00:43:50 No.7384680 Report

Quoted By: >>7384697

>>7384651
Excellent as always

Anonymous

Anonymous Sat 20 May 2023 00:44:38 No.7384682 Report

Quoted By: >>7384693 >>7384695

>>7384656
wrong tags can massively screw things up
not so much new tags so much as incorrect tags
if you don't tag something, it'll appear with or without those tags, for example tagging shirt but not necktie will show the necktie anyway (assuming your data has both variations)
if you tag something incorrectly, then trying to adjust it will be less flexible, for example tagging "large breasts" instead of "medium breasts" will cause breast adjustments to not work as you'd expect.

Anonymous

Anonymous Sat 20 May 2023 00:47:29 No.7384686 Report

Quoted By: >>7384693

>>7384677
i'm pretty sure that a bunch of people have talked about it though

a few one off incorrect captions won't be an issue, but they have a detrimental effect if they're actually related to your character. having your character's captions be incorrect or inconsistent can make it harder for the AI to figure out what is or isn't in a training image.

you can also use tags proactively to fix certain aspects of a lora. if there's a consistent element of a character design that artists get wrong you can add a caption to those images describing it and have those undesired elements appear less in the final lora

a lot of results will vary and be inconsistent across datasets because you're effectively fine tuning part of an existing model, and the existing weights may have unpredictable effects on your captions. probably most of my time spent training is just retraining the same lora with different captions over and over to see which ones work best.

Anonymous

Anonymous Sat 20 May 2023 00:50:41 No.7384690 Report

Quoted By:

fucking love vex porn
but i gotta wonder where the line is
vex is arguably more /h/ than /trash/ and ankha is arguably more /trash/ than /h/

Anonymous

View Same Google ImgOps iqdb SauceNAO 01021-2193338622.png, 1MiB, 1024x1536

Anonymous Sat 20 May 2023 00:52:43 No.7384692 Report

Quoted By: >>7384698 >>7384704

>>7384661
I don't watch her at all but here's a quick one for you

Anonymous

Anonymous Sat 20 May 2023 00:53:09 No.7384693 Report

Quoted By:

>>7384682
I think its the same on a fuckup scale, if anything tagging something wrong should be less of an issue unless it repeats alot throughout the dataset - its pretty hard to overpower the model's understanding

>>7384686
>probably most of my time spent training is just retraining the same lora with different captions over and over to see which ones work best.
can you give an example of this? It shouldnt matter what caption you use aside from consolidating to most accurate

Anonymous

View Same Google ImgOps iqdb SauceNAO file.png, 13KiB, 830x83

Anonymous Sat 20 May 2023 00:54:17 No.7384695 Report

Quoted By:

>>7384682
>sizes
yeah im just gonna let AI sort it out, fucking hell

Anonymous

Anonymous Sat 20 May 2023 00:57:14 No.7384697 Report

Quoted By: >>7384705

>>7384680
Not bad yourself there.

Anonymous

View Same Google ImgOps iqdb SauceNAO 01023-1955582341.png, 1MiB, 1024x1536

Anonymous Sat 20 May 2023 00:57:22 No.7384698 Report

Quoted By: >>7384704

>>7384661
>>7384692
One more for her

Anonymous

Anonymous Sat 20 May 2023 01:03:44 No.7384704 Report

Quoted By:

>>7384692
>>7384698
neither do I cause vtubers are aids, but thanks. Haato a cute.

Anonymous

View Same Google ImgOps iqdb SauceNAO catbox_ib3f9m.png, 1MiB, 1024x1024

Anonymous Sat 20 May 2023 01:04:04 No.7384705 Report

Quoted By:

>>7384697
Thanks

Anonymous

View Same Google ImgOps iqdb SauceNAO 00917-2473371365-best quality, ( (...).png, 577KiB, 656x960

Anonymous Sat 20 May 2023 01:08:24 No.7384709 Report

Quoted By: >>7384725

Anonymous

View Same Google ImgOps iqdb SauceNAO 01037-3265567992.png, 1MiB, 1024x1536

Anonymous Sat 20 May 2023 01:10:17 No.7384712 Report

Quoted By:

Anonymous

View Same Google ImgOps iqdb SauceNAO 00128-881007884.png, 1MiB, 1152x1152

Anonymous Sat 20 May 2023 01:12:37 No.7384713 Report

Quoted By: >>7384718 >>7384723 >>7384922

I feel I got some good improvement again, but I'm starting to hit the wall of try random shit. Loras are really fucking cool.

>>7384657
I have 10gb, is that enough to train at 768 and still do other stuff comfortably if I optimize?

Anonymous

View Same Google ImgOps iqdb SauceNAO 00092-122647871.png, 1MiB, 768x1152

Anonymous Sat 20 May 2023 01:14:57 No.7384717 Report

Quoted By:

Anonymous

Anonymous Sat 20 May 2023 01:16:39 No.7384718 Report

Quoted By:

>>7384713
>other stuff comfortably
Depends what stuff you mean. If its watching stuff or something that barely uses gpu then yeah maybe.
If its gaming then pretty sure thats a no but I guess you can check. Worst case it'll error in the middle because your game will demand more vram at some point.
At 8GB I can only run batch1 on 768, not sure how much vram does it take exactly but I'd wager around 6-7. Am lucky to have an integrated gpu to fallback on for some light gaming

Anonymous

View Same Google ImgOps iqdb SauceNAO 01058-1563546908.png, 2MiB, 1024x1536

Anonymous Sat 20 May 2023 01:20:05 No.7384719 Report

Quoted By:

Anonymous

View Same Google ImgOps iqdb SauceNAO 00933-3194549939-best quality, ( (...).png, 670KiB, 656x960

Anonymous Sat 20 May 2023 01:25:11 No.7384720 Report

Quoted By: >>7384928

Anonymous

View Same Google ImgOps iqdb SauceNAO 01743-3306289712.png, 2MiB, 1536x1024

Anonymous Sat 20 May 2023 01:27:44 No.7384722 Report

Quoted By:

>>7384672
This is the bug report for this error on webui:
https://github.com/AUTOMATIC1111/stable-diffusion-webui/issues/9341

Perhaps the only line you need to add is this, but I haven't tried it:
pip install torch==2.0.1+cu118 --extra-index-url https://download.pytorch.org/whl/cu118

Anonymous

Anonymous Sat 20 May 2023 01:27:59 No.7384723 Report

Quoted By:

>>7384713
turn on gradient checkpointing bro

Anonymous

Anonymous Sat 20 May 2023 01:28:53 No.7384724 Report

Quoted By:

Why is it so hard for arms not to turn into dicks?

Anonymous

Anonymous Sat 20 May 2023 01:29:43 No.7384725 Report

Quoted By: >>7384732 >>7384735 >>7384743 >>7384746

>>7384709
more rosemi please, with bigger breasts please

Anonymous

Anonymous Sat 20 May 2023 01:36:32 No.7384730 Report

Quoted By: >>7384744

>>7384661
>more 1girl
whoa
impressive

Anonymous

View Same Google ImgOps iqdb SauceNAO 00248-2272897547-nsfw, scan, (tr (...).png, 2MiB, 1024x1536

Anonymous Sat 20 May 2023 01:37:22 No.7384731 Report

Quoted By: >>7384740 >>7384921

Mommy erotic.

Anonymous

View Same Google ImgOps iqdb SauceNAO 00940-3664054072-best quality, ( (...).png, 763KiB, 960x656

Anonymous Sat 20 May 2023 01:37:24 No.7384732 Report

Quoted By: >>7384928

>>7384725

Anonymous

Anonymous Sat 20 May 2023 01:39:46 No.7384735 Report

Quoted By:

>>7384725
wosemi is not lewd

Anonymous

Anonymous Sat 20 May 2023 01:43:36 No.7384740 Report

Quoted By: >>7384745

>>7384731
grabable fattie

Anonymous

View Same Google ImgOps iqdb SauceNAO 00948-1497848912-best quality, ( (...).png, 648KiB, 656x960

Anonymous Sat 20 May 2023 01:44:27 No.7384742 Report

Quoted By:

Anonymous

View Same Google ImgOps iqdb SauceNAO 01092-2808267927.png, 1MiB, 1024x1536

Anonymous Sat 20 May 2023 01:45:12 No.7384743 Report

Quoted By:

>>7384725
Sure why not

Anonymous

Anonymous Sat 20 May 2023 01:46:55 No.7384744 Report

Quoted By:

>>7384730
Anything else is too hard to make

Anonymous

Anonymous Sat 20 May 2023 01:47:26 No.7384745 Report

Quoted By:

>>7384740
(((HEALTHY)))

Anonymous

View Same Google ImgOps iqdb SauceNAO 00952-1482866318-best quality, ( (...).png, 641KiB, 656x960

Anonymous Sat 20 May 2023 01:49:00 No.7384746 Report

Quoted By: >>7384776

>>7384725
sorry I'm moving between different characters

Anonymous

View Same Google ImgOps iqdb SauceNAO 00150-3267829432.png, 2MiB, 1280x1536

Anonymous Sat 20 May 2023 02:01:07 No.7384754 Report

Quoted By:

Anonymous

View Same Google ImgOps iqdb SauceNAO 00151-1171490668.png, 449KiB, 512x768

Anonymous Sat 20 May 2023 02:15:38 No.7384764 Report

Quoted By: >>7384769 >>7384774 >>7384851 >>7384922

How should I tag these kinds of breast out images in the dataset. Also just the nipples poking out, generating a lot of those.

Anonymous

Anonymous Sat 20 May 2023 02:24:22 No.7384769 Report

Quoted By:

>>7384764
bad anatomy

Anonymous

Anonymous Sat 20 May 2023 02:31:03 No.7384774 Report

Quoted By:

>>7384764
tanaka_hitomi

Anonymous

Anonymous Sat 20 May 2023 02:32:03 No.7384776 Report

Quoted By:

>>7384746
based flattie fixer

Anonymous

Anonymous Sat 20 May 2023 02:36:58 No.7384778 Report

Quoted By: >>7384787 >>7384789

I switched to an nvme and literally soigaped at how instantly I can switch models. Now if only I had something faster than a 3070

Anonymous

Anonymous Sat 20 May 2023 02:45:01 No.7384786 Report

Quoted By:

i've been gone 5 days. did i miss any good shared images or loras?

Anonymous

View Same Google ImgOps iqdb SauceNAO 2389927619.png, 1MiB, 880x1280

Anonymous Sat 20 May 2023 02:46:00 No.7384787 Report

Quoted By: >>7384789 >>7386272

>>7384778
My NVME seemed like a waste of money until the last couple of months with AI art, now I'm so glad I bought one.

Anonymous

Anonymous Sat 20 May 2023 02:48:42 No.7384789 Report

Quoted By: >>7384790

>>7384778
>>7384787
you guys get the 990 pro 2tb for $140?

Anonymous

Anonymous Sat 20 May 2023 02:54:43 No.7384790 Report

Quoted By:

>>7384789
980 pro 2TB and mine switches between models in 2 seconds.

Anonymous

View Same Google ImgOps iqdb SauceNAO 00160-414833652.png, 2MiB, 1280x1536

Anonymous Sat 20 May 2023 02:56:37 No.7384793 Report

Quoted By:

>>7384548
Sorry, totally forgot about it.

Here it is, probably gonna upload other mixes later.

https://huggingface.co/BGLuck/BlockMergeWeightedMixes/blob/main/peculiar_xxmix.safetensors

Named the mix like that just because one of the models used there IoI.

Anonymous

View Same Google ImgOps iqdb SauceNAO 00025-3741596476.jpg, 1MiB, 1024x1440

Anonymous Sat 20 May 2023 03:24:12 No.7384820 Report

Quoted By: >>7384833

Something has gone awry, but I can't put my finger on it.

Anonymous

Anonymous Sat 20 May 2023 03:35:18 No.7384833 Report

Quoted By:

>>7384820
Is this a great secret of the oppai loli?

Anonymous

Anonymous Sat 20 May 2023 03:37:54 No.7384834 Report

Quoted By:

>>7384579
Damn she's taking it inside the womb and bulging through the stomach without as much as a flinch. And my dude's benis is as thick and girthy as his legs.

Anonymous

Anonymous Sat 20 May 2023 03:48:12 No.7384841 Report

Quoted By:

Are there any new optimizations I should do for a fresh install of auto1111? I haven't prompted in a few months so I'm wondering if there's anything I can do for my 3070ti or if I just have to cope.

Anonymous

View Same Google ImgOps iqdb SauceNAO file.png, 98KiB, 1249x764

Anonymous Sat 20 May 2023 03:52:00 No.7384843 Report

Quoted By: >>7385028

Huh you now can use a different sampler when upscaling on txt2img, nice

(dev branch, maybe it is on the master one too?

Anonymous

Anonymous Sat 20 May 2023 04:02:57 No.7384851 Report

Quoted By: >>7384857

>>7384764
catbox? I forget what lora that is

Anonymous

Anonymous Sat 20 May 2023 04:05:35 No.7384857 Report

Quoted By: >>7384875

>>7384851
it's my uni lora, no where near ready, you're probably thinking about alexi. and don't ask for my shitty test prompt when the stuff I put effort into gets no (You)'s, you're killing me

Anonymous

View Same Google ImgOps iqdb SauceNAO zcxvqxta.png, 2MiB, 1024x1536

Anonymous Sat 20 May 2023 04:17:59 No.7384875 Report

Quoted By: >>7384880 >>7384884 >>7387202

>>7384857
what is a (You)?

Anonymous

Anonymous Sat 20 May 2023 04:21:34 No.7384880 Report

Quoted By: >>7384929 >>7384953 >>7385173

>>7384875
I have the same issue as you, with the hair turning blue behind the neck

Anonymous

Anonymous Sat 20 May 2023 04:23:03 No.7384884 Report

Quoted By:

>>7384875
jesus fucking christ

Anonymous

View Same Google ImgOps iqdb SauceNAO supper.png, 3MiB, 1832x1152

Anonymous Sat 20 May 2023 04:24:32 No.7384887 Report

Quoted By: >>7385076

Anonymous

View Same Google ImgOps iqdb SauceNAO 00176-2860006718.png, 391KiB, 600x600

Anonymous Sat 20 May 2023 04:30:46 No.7384894 Report

Quoted By: >>7384921 >>7385022 >>7385055

Anonymous

View Same Google ImgOps iqdb SauceNAO 00011-417781068.png, 2MiB, 1280x1280

Anonymous Sat 20 May 2023 04:49:34 No.7384917 Report

Quoted By: >>7385353

Anonymous

Anonymous Sat 20 May 2023 04:54:50 No.7384920 Report

Quoted By:

>>7384406
Thanks
1girl, portrait, hakurei reimu, nude, medium breasts, red bow, flat colors, limited palette, sketch, covering crotch <lyco:coveringV2:0.6>
Negative prompt: (worst quality, low quality:1.2), loli, flat chest, small breasts, jpeg artifacts
AddNet Model 1: HISHA_KAN_MOKO_V1-08(655031d02688)
LoRA, AddNet Model 2: artist-gomennasai(d72673d6cdb3)
LoRA, AddNet Model 3: add_detail(7c6bad76eb54)

In a way it's a fail of the lora because the gen didn't explicitly show her covering crotch, but the pose came out nice anyway

Anonymous

Anonymous Sat 20 May 2023 04:55:10 No.7384921 Report

Quoted By: >>7385313

>>7384731
>>7384894
Catbox please?

Anonymous

Anonymous Sat 20 May 2023 04:55:27 No.7384922 Report

Quoted By: >>7384940

>>7384533
>>7384713
>>7384764
Based, more Uni please.

Anonymous

View Same Google ImgOps iqdb SauceNAO bd3eb5ad5776b3f5491e8c3bf9dd50fc (...).png, 1MiB, 1024x1536

Anonymous Sat 20 May 2023 05:01:21 No.7384928 Report

Quoted By: >>7385384

>>7384720
>>7384732
Nice. Catbox, onegai?

Anonymous

Anonymous Sat 20 May 2023 05:02:46 No.7384929 Report

Quoted By:

>>7384880
What model and LoRAs do you use?

Anonymous

View Same Google ImgOps iqdb SauceNAO 30595514.png, 498KiB, 512x768

Anonymous Sat 20 May 2023 05:04:23 No.7384931 Report

Quoted By: >>7384934

Dark Skinned Femboy

Anonymous

View Same Google ImgOps iqdb SauceNAO 00043-382540491.png, 2MiB, 1280x1536

Anonymous Sat 20 May 2023 05:05:20 No.7384932 Report

Quoted By:

Anonymous

Anonymous Sat 20 May 2023 05:09:45 No.7384934 Report

Quoted By:

>>7384931
>boy

Anonymous

View Same Google ImgOps iqdb SauceNAO 00055-1376074536.png, 903KiB, 1024x1024

Anonymous Sat 20 May 2023 05:16:12 No.7384940 Report

Quoted By: >>7385144 >>7385243

>>7384922

Anonymous

Anonymous Sat 20 May 2023 05:34:56 No.7384951 Report

Quoted By:

>>7383633
i mean, do you still type them? i just have them autofilled with the autofinish plugin

Anonymous

View Same Google ImgOps iqdb SauceNAO 02030-2093665496.png, 2MiB, 1536x1024

Anonymous Sat 20 May 2023 05:35:39 No.7384953 Report

Quoted By: >>7385046

>>7384880
Put "multicolored hair" in the negative prompt.

Anonymous

View Same Google ImgOps iqdb SauceNAO 00018-715880163.png, 2MiB, 1536x1024

Anonymous Sat 20 May 2023 05:38:14 No.7384956 Report

Quoted By: >>7384958 >>7385106

Anonymous

Anonymous Sat 20 May 2023 05:41:24 No.7384958 Report

Quoted By: >>7384961

>>7384956
this one's nice, catbox?

Anonymous

Anonymous Sat 20 May 2023 05:45:20 No.7384961 Report

Quoted By: >>7384962 >>7385106 >>7385354

>>7384958
https://files.catbox.moe/vag7g8.png
Using Regional prompter. Minor tweak in photoshop to fix the hand. img2img upscale and inpainting for the faces and tongues.

Anonymous

Anonymous Sat 20 May 2023 05:47:46 No.7384962 Report

Quoted By: >>7384963 >>7384971 >>7384979

>>7384961
>mixProV4_v4
this model any good or just ridiculously overfit civitshit?

Anonymous

Anonymous Sat 20 May 2023 05:50:03 No.7384963 Report

Quoted By: >>7384965

>>7384962
use nai if you don't like overfit

Anonymous

Anonymous Sat 20 May 2023 05:52:14 No.7384965 Report

Quoted By:

>>7384963
no? AOM is the right kind of overfit. your garbage model with a billion loras mixed into it is not

Anonymous

Anonymous Sat 20 May 2023 05:57:14 No.7384971 Report

Quoted By:

>>7384962
I like it a lot. It's very amazing. One of my favourite models.

Anonymous

Anonymous Sat 20 May 2023 06:05:43 No.7384979 Report

Quoted By:

>>7384962
It is pretty good IMO, but I like v3 better than V4 or 4.5

Anonymous

Anonymous Sat 20 May 2023 06:41:01 No.7385010 Report

Quoted By: >>7385012 >>7385021 >>7385024

I want to clean out my lora hoard, what's the best tool to find files with the same hash and delete one of them?

Anonymous

Anonymous Sat 20 May 2023 06:41:07 No.7385011 Report

Quoted By:

Don't think it's been shared here, abubu made an interesting breakdown of one of their AI-assissted works recently. There's a few more on their Note page. https://note.com/abubu_nounanka/n/na9c13d38c445

Anonymous

Anonymous Sat 20 May 2023 06:42:08 No.7385012 Report

Quoted By: >>7385018

>>7385010
https://github.com/qarmin/czkawka

Anonymous

Anonymous Sat 20 May 2023 06:53:58 No.7385018 Report

Quoted By:

>>7385012
Thanks, looks useful

Anonymous

View Same Google ImgOps iqdb SauceNAO 00096-3264223809.png, 1MiB, 1280x1536

Anonymous Sat 20 May 2023 06:57:27 No.7385020 Report

Quoted By:

Anonymous

Anonymous Sat 20 May 2023 07:01:06 No.7385021 Report

Quoted By: >>7385035 >>7385059

>>7385010
That reminds me do you guys keep LoRAs you rarely use/used once?

Anonymous

Anonymous Sat 20 May 2023 07:01:16 No.7385022 Report

Quoted By:

>>7384894
i want to live in those nipples

Anonymous

Anonymous Sat 20 May 2023 07:03:03 No.7385024 Report

Quoted By: >>7385395

>>7385010
you could very likely find a github gist with powershell code (or bash if linux) to do this. too lazy to think what the linux commands would be but ask chatgpt

Anonymous

View Same Google ImgOps iqdb SauceNAO 00101-198098288.png, 2MiB, 1280x1536

Anonymous Sat 20 May 2023 07:04:11 No.7385027 Report

Quoted By:

Anonymous

Anonymous Sat 20 May 2023 07:04:17 No.7385028 Report

Quoted By:

>>7384843
>Sampling steps: 80

Anonymous

View Same Google ImgOps iqdb SauceNAO 00055-2863802187.png, 2MiB, 1536x1024

Anonymous Sat 20 May 2023 07:08:35 No.7385031 Report

Quoted By: >>7385048 >>7385077

Anonymous

Anonymous Sat 20 May 2023 07:10:52 No.7385035 Report

Quoted By:

>>7385021
yes. I have a problem.

Anonymous

View Same Google ImgOps iqdb SauceNAO 00565-3403591020.png, 1MiB, 1024x1536

Anonymous Sat 20 May 2023 07:16:30 No.7385042 Report

Quoted By: >>7385044

I don't even know what's going on in this one.

Anonymous

Anonymous Sat 20 May 2023 07:19:00 No.7385044 Report

Quoted By:

>>7385042
It's simple. She's having a wet dream and taking it out on the curtain.

Anonymous

Anonymous Sat 20 May 2023 07:21:32 No.7385046 Report

Quoted By: >>7385063

>>7384953
catbox? I've been trying for space bg and it always looks fucked

Anonymous

Anonymous Sat 20 May 2023 07:22:40 No.7385048 Report

Quoted By:

>>7385031
taking right

Anonymous

Anonymous Sat 20 May 2023 07:28:00 No.7385055 Report

Quoted By:

>>7384894
catbox?

Anonymous

Anonymous Sat 20 May 2023 07:32:28 No.7385059 Report

Quoted By:

>>7385021
I have ~850 loras saved after cleaning up and I used like 50 at best.

Anonymous

View Same Google ImgOps iqdb SauceNAO 02044-4050391160.png, 2MiB, 1536x1024

Anonymous Sat 20 May 2023 07:34:19 No.7385063 Report

Quoted By:

>>7385046
https://files.catbox.moe/s4ngkw.png

Anonymous

View Same Google ImgOps iqdb SauceNAO 00141-1922330904.png, 2MiB, 1280x1536

Anonymous Sat 20 May 2023 07:35:02 No.7385064 Report

Quoted By:

Anonymous

Anonymous Sat 20 May 2023 07:48:32 No.7385073 Report

Quoted By: >>7385078 >>7385261

do you have any other lora training guide that isn't using LoRA_Easy_Training_Scripts? torch devs deleted nightly 2.1 wheel associated to it, if i use any other torch wheel - it makes xformers wheel useless aswel and training throws "no kernel image is available for execution on the device" error

or, are there any prebuilt xformers wheel for torch 2.1 to download?

Anonymous

Anonymous Sat 20 May 2023 07:56:16 No.7385076 Report

Quoted By:

>>7384887
>the true ending as seem by Fuminori
wholesome.
did you copy paste the goo and inpainted or was this prompted for? iirc it was tagged as "slime" in the set

Anonymous

Anonymous Sat 20 May 2023 07:58:38 No.7385077 Report

Quoted By: >>7385095

>>7385031
Do you have a pixiv for something I can follow

Anonymous

Anonymous Sat 20 May 2023 07:58:47 No.7385078 Report

Quoted By: >>7385090

>>7385073
cant run it on torch 2.1 anymore cause it cant install? am I getting it right? cant you just coy paste someone else's install folder to fix that? or just run it on torch 2.0? diff should be minimal anyway

Anonymous

Anonymous Sat 20 May 2023 08:00:28 No.7385082 Report

Quoted By:

>>7384057
thanks dude, yeah wanted to see which samplers you use and if you did original or fill, also the denoise

Anonymous

Anonymous Sat 20 May 2023 08:03:22 No.7385084 Report

Quoted By: >>7386822

>>7383586
where can i find instant loss lora

Anonymous

Anonymous Sat 20 May 2023 08:08:42 No.7385090 Report

Quoted By: >>7385099 >>7385114

>>7385078
at the moment if you choose to install torch 2.1 it will skip torch installation because it cannot download nightly build because it was deleted
>cant you just coy paste someone else's install folder to fix that?
if anon used this https://rentry.org/LazyTrainingGuide guide that asks you to use torch 2.1 then yes
>or just run it on torch 2.0?
for some reason training bat tries to use CPU if im using torch 2.0, seems like it cannot find my GPU

Anonymous

Anonymous Sat 20 May 2023 08:11:19 No.7385093 Report

Quoted By:

...
I just trained stuff overnight but forgot to remove all underscores.
its so over

Anonymous

Anonymous Sat 20 May 2023 08:13:09 No.7385095 Report

Quoted By:

>>7385077
Nope.

Anonymous

Anonymous Sat 20 May 2023 08:13:43 No.7385097 Report

Quoted By:

>>7384478
no I think's he right, some models just drastically inpaint better because it works differently. There was a reddit thread that explained how to turn any model into an inpaint model

Anonymous

Anonymous Sat 20 May 2023 08:14:07 No.7385098 Report

Quoted By: >>7385111

So where do you post these images, social media stuff?
This is a fun hobby and if I can get more enjoyment out of it from many (You)s, I'd do it.

Anonymous

Anonymous Sat 20 May 2023 08:16:31 No.7385099 Report

Quoted By:

>>7385090
might have to open an issue in kohya's repo or easyscripts then

Anonymous

Anonymous Sat 20 May 2023 08:22:34 No.7385101 Report

Quoted By: >>7385385 >>7385398

>>7384457
not that anon but can you explain the workflow in a bit more detail, I know the ui it's just the naming of the settings which I don't get.

Or if you know a good yt guide, the vids that I've seen were so different from each other to the point where they were just guessing stuff or contradicting each other which made people even more confused

Anonymous

Anonymous Sat 20 May 2023 08:23:46 No.7385102 Report

Quoted By:

ay so i downloaded a new model and placed it in the directory, but when i refresh and restart the ui it dosent appear. what do?

Anonymous

Anonymous Sat 20 May 2023 08:27:08 No.7385106 Report

Quoted By:

>>7384956
>>7384961
You upscale in img2img? looks good

Anonymous

Anonymous Sat 20 May 2023 08:38:58 No.7385111 Report

Quoted By:

>>7385098
pixiv, twitter

Anonymous

View Same Google ImgOps iqdb SauceNAO 11296-1249395640-masterpiece, be (...).png, 1MiB, 1024x1280

Anonymous Sat 20 May 2023 08:39:43 No.7385112 Report

Quoted By:

Anonymous

Anonymous Sat 20 May 2023 08:43:17 No.7385114 Report

Quoted By: >>7385121

>>7385090
update: got same "no kernel image is available for execution on the device" error on torch 2.0
will try to reinstall all packages and python

Anonymous

View Same Google ImgOps iqdb SauceNAO 00472-3143803229.png, 469KiB, 600x600

Anonymous Sat 20 May 2023 08:45:44 No.7385115 Report

Quoted By:

Gurren lagann tag gives some nice color schemes

Anonymous

View Same Google ImgOps iqdb SauceNAO 00384-621492445.png, 372KiB, 512x512

Anonymous Sat 20 May 2023 08:46:39 No.7385116 Report

Quoted By:

Anonymous

View Same Google ImgOps iqdb SauceNAO 00180-879461203.png, 305KiB, 512x512

Anonymous Sat 20 May 2023 08:47:36 No.7385117 Report

Quoted By:

Anonymous

Anonymous Sat 20 May 2023 08:58:06 No.7385121 Report

Quoted By: >>7385209

>>7385114
iirc one of the guides/faq mentioned not detecting your GPU, might be worth checking >>7383628
or /g/ thread links

Anonymous

View Same Google ImgOps iqdb SauceNAO catbox_nq4tdi.png, 1MiB, 1024x1024

Anonymous Sat 20 May 2023 09:07:58 No.7385129 Report

Quoted By: >>7385133

Anonymous

View Same Google ImgOps iqdb SauceNAO catbox_mo911a.png, 1MiB, 1024x1024

Anonymous Sat 20 May 2023 09:12:39 No.7385133 Report

Quoted By: >>7385142

>>7385129

Anonymous

Anonymous Sat 20 May 2023 09:22:05 No.7385141 Report

Quoted By: >>7385177

does anyone have a lora that has bras and underwear?

Anonymous

View Same Google ImgOps iqdb SauceNAO catbox_f0j3kf.png, 1MiB, 1024x1024

Anonymous Sat 20 May 2023 09:22:08 No.7385142 Report

Quoted By: >>7385195

>>7385133

Anonymous

View Same Google ImgOps iqdb SauceNAO image_2023-05-20_132157835.png, 1MiB, 1024x1536

Anonymous Sat 20 May 2023 09:22:21 No.7385144 Report

Quoted By: >>7385243 >>7385737

>>7384940
alexianon here, have to say your work is commendable

Anonymous

Anonymous Sat 20 May 2023 09:31:18 No.7385148 Report

Quoted By: >>7385196

Is there a lewd/hentai stable diffusion discord to exchange pics? Thank you.

Anonymous

Anonymous Sat 20 May 2023 09:36:44 No.7385153 Report

Quoted By: >>7385181

>all these godawful overbaked style loras on civit
scrolling through this site is so shit too, it's impossible to browse through everything due to the fact it has no pages so you can never pick up around where you left off.

Anonymous

Anonymous Sat 20 May 2023 09:59:16 No.7385173 Report

Quoted By:

>>7384880
just put blue hair in negs

Anonymous

Anonymous Sat 20 May 2023 10:05:14 No.7385177 Report

Quoted By: >>7385184 >>7385939

>>7385141
yeah

Anonymous

Anonymous Sat 20 May 2023 10:09:08 No.7385181 Report

Quoted By:

>>7385153
>hes just browsing style loras with no goal in mind
lol
go browse art on boorus and look up the artist instead dumbass

Anonymous

Anonymous Sat 20 May 2023 10:12:16 No.7385184 Report

Quoted By: >>7385189

>>7385177
can you share it?

Anonymous

Anonymous Sat 20 May 2023 10:13:03 No.7385185 Report

Quoted By: >>7385191

Can anyone share a 1girl lora please?

Anonymous

Anonymous Sat 20 May 2023 10:20:29 No.7385189 Report

Quoted By: >>7385204

>>7385184
why you would need this in the 1st place

they are all in the op

Anonymous

Anonymous Sat 20 May 2023 10:22:07 No.7385191 Report

Quoted By:

>>7385185
its called ((((solo))))

Anonymous

Anonymous Sat 20 May 2023 10:26:47 No.7385195 Report

Quoted By:

>>7385142
And they say AI images aren't art

Anonymous

Anonymous Sat 20 May 2023 10:27:50 No.7385196 Report

Quoted By: >>7385199

>>7385148
Unstable Diffusion is your best bet

Anonymous

Anonymous Sat 20 May 2023 10:29:31 No.7385199 Report

Quoted By:

>>7385196
Thank you kind anon

Anonymous

Anonymous Sat 20 May 2023 10:33:25 No.7385204 Report

Quoted By: >>7385211

>>7385189
That's the thing, i can't find one. At least not in the gitgud.io link. CivitAI doesn't help either.

Anonymous

Anonymous Sat 20 May 2023 10:40:39 No.7385209 Report

Quoted By: >>7385255

>>7385121
and kohya_ss gui doesn't recognize its own xformers installed through guide, and even if i disable them - it throws some "The paging file is too small for this operation to complete."

im too retarded to make second lora i guess

Anonymous

Anonymous Sat 20 May 2023 10:45:20 No.7385211 Report

Quoted By: >>7385230 >>7385933

>>7385204
>crotchless panties
>bra visible through clothes
>see-through silhouette
all in the op

if you are having trouble just prompting underwear then it's a skill issue and you should probably avail yourself of https://danbooru.donmai.us/wiki_pages/tag_group%3Asexual_attire

Anonymous

View Same Google ImgOps iqdb SauceNAO 110db50f32143191f8ecc90153e7190f.png, 130KiB, 2412x1103

Anonymous Sat 20 May 2023 11:06:19 No.7385228 Report

Quoted By:

You should include popular extensions in the op.
I just found out about this one, pic related:
https://github.com/GeorgLegato/stable-diffusion-webui-artists-ui-tweak

Anonymous

Anonymous Sat 20 May 2023 11:07:19 No.7385230 Report

Quoted By: >>7385234

>>7385211
>guy asks for lora
>it's in the op
>can't find it in op
>it's in the op bro, here's tags you can use (not a lora)
lol

Anonymous

Anonymous Sat 20 May 2023 11:14:07 No.7385234 Report

Quoted By: >>7385239

>>7385230
ctrl-f those phrases in the op fucking moron

Anonymous

Anonymous Sat 20 May 2023 11:19:33 No.7385239 Report

Quoted By: >>7385242 >>7385249 >>7385251

>>7385234
>fucking moron
hey i'm not the one giving completely unrelated answers to questions. guy said he couldn't find any in the resources listed, and you just called him shit and started linking tags to flex on him for being able to prompt underwear when he was asking for a lora, probably for specific stuff? lmao get real loser

Anonymous

View Same Google ImgOps iqdb SauceNAO aieeee.webm, 1MiB, 1316x834

Anonymous Sat 20 May 2023 11:21:46 No.7385242 Report

Quoted By: >>7385244

>>7385239
AIIIEEEEEEEEEEEEEE NOOO NOT CTRL-F NOO NOO NOOOOOOOOOOOOOOOOOO

Anonymous

Anonymous Sat 20 May 2023 11:23:12 No.7385243 Report

Quoted By:

>>7384940
>>7385144
Based anons.

Anonymous

Anonymous Sat 20 May 2023 11:24:45 No.7385244 Report

Quoted By: >>7385251

>>7385242
>autistic screeching
lol

Anonymous

View Same Google ImgOps iqdb SauceNAO 00009-1989624664.png, 1MiB, 1024x1536

Anonymous Sat 20 May 2023 11:26:55 No.7385246 Report

Quoted By: >>7385248 >>7385270

https://mega.nz/folder/ctl1FYoK#BYlpywutnH4psbEcxsbLLg
added lm (legoman)

Anonymous

View Same Google ImgOps iqdb SauceNAO 00016-3882055313.png, 973KiB, 1024x1536

Anonymous Sat 20 May 2023 11:28:04 No.7385248 Report

Quoted By:

>>7385246
i guess i should add curvy, mature female, and general milf tags bring out his style more

Anonymous

Anonymous Sat 20 May 2023 11:28:17 No.7385249 Report

Quoted By: >>7385250

>>7385239
how do people like you not forget how to breath?

Anonymous

Anonymous Sat 20 May 2023 11:29:14 No.7385250 Report

Quoted By: >>7385256

>>7385249
>new ip
trying too hard

Anonymous

Anonymous Sat 20 May 2023 11:29:32 No.7385251 Report

Quoted By:

>>7385244
>>7385239
well he is right
how do you go from
>checking gitgud.io link
to
>can't find one
did he just look at the wall of links and text and go... "yeah ill just ask in the thread and have people find it for me :)" like how far back do you have to be not to develop ctrl-f skills? is he using some obscure browser from 1995 that doesn't support the feature?

Anonymous

View Same Google ImgOps iqdb SauceNAO 00019-4172888044.png, 3MiB, 1280x1920

Anonymous Sat 20 May 2023 11:30:24 No.7385252 Report

Quoted By:

schizo hours atm. damn shame

Anonymous

View Same Google ImgOps iqdb SauceNAO xyz_grid-0012-2614686574-solo, N (...).jpg, 332KiB, 4000x409

Anonymous Sat 20 May 2023 11:31:21 No.7385255 Report

Quoted By: >>7385259 >>7385266

I was talking about tags yesterday, and this is the weird shit I meant. For some reason the 0.3 WD run knows how to output stairs properly, but the 0.15 one that was pruned through doesn't, despite me checking if there's any images mistagged with "stairs". The only image with this tag is https://files.catbox.moe/omrzol.jpg but that's not even a mistake. And yes it was tested on a few grids with the same outcome.
tl;dr is tagging everything actually damaging to the lora and the old advice of "mistagged things will appear" bullshit in a way?

>>7385209
paging file can be raised through your system settings, google it, its really easy.
Unsure why you've so many xformer issues though. You can try old torch through easyscripts? It'll just be slower

Anonymous

Anonymous Sat 20 May 2023 11:32:24 No.7385256 Report

Quoted By: >>7385257

>>7385250
Dude, I just woke up and first thing I read is bullshit from retards to stupid to read.
I get it you embarrassed yourself and are trying to play the tough dude now to play it off. Just take your lora and do something useful.

Anonymous

Anonymous Sat 20 May 2023 11:33:52 No.7385257 Report

Quoted By: >>7385258 >>7385259

>>7385256
>trying to play the tough dude
lmao who takes an imageboard this seriously. kys

Anonymous

Anonymous Sat 20 May 2023 11:35:21 No.7385258 Report

Quoted By:

>>7385257
fuck off zoomer

Anonymous

Anonymous Sat 20 May 2023 11:35:40 No.7385259 Report

Quoted By: >>7385262

>>7385255
like i was saying yesterday i think 0.15 is just too damn low

>>7385257
what kind of moron can't use ctrl-f

Anonymous

Anonymous Sat 20 May 2023 11:39:34 No.7385261 Report

Quoted By:

>>7385073
>do you have any other lora training guide that isn't using LoRA_Easy_Training_Scripts?
i love how often i read in this and the last few threads about how shit the new op is. it's poetic

Anonymous

Anonymous Sat 20 May 2023 11:39:43 No.7385262 Report

Quoted By: >>7385267 >>7385272

>>7385259
what does 0.15 being too low have to do with this though? its tagged correctly.
If there are things that should not be tagged, that's an entirely different issue

Anonymous

Anonymous Sat 20 May 2023 11:44:29 No.7385266 Report

Quoted By: >>7385267

>>7385255
>You can try old torch through easyscripts?
2.0 or 1.12.1?
because 2.0 throws "no kernel image is available for execution on the device" error and it seems like nobody else ever got it

Anonymous

Anonymous Sat 20 May 2023 11:45:31 No.7385267 Report

Quoted By: >>7385284

>>7385262
Eh. Kind of making me realize how bad game CGs are for this, as they're usually more cluttered with random background stuff than booru art..

>>7385266
meant 1.12.1
open an issue about 2.0 or look through sd-scripts issues if you havent yet

Anonymous

View Same Google ImgOps iqdb SauceNAO 01153-1706611387.png, 394KiB, 1048x1568

Anonymous Sat 20 May 2023 11:47:41 No.7385270 Report

Quoted By:

>>7385246
neat

Anonymous

Anonymous Sat 20 May 2023 11:51:19 No.7385272 Report

Quoted By: >>7385275

>>7385262
well, the prompt it's training on is just a normal generation, it'll be less clear and more filled with stuff the longer the prompt is. and order matters too just like normal prompting. there is absolutely a difference between 25 tokens of what you're training on + 15 tokens of random stuff vs 25 tokens of what you're training on + 150 tokens of random stuff

Anonymous

Anonymous Sat 20 May 2023 11:56:36 No.7385275 Report

Quoted By: >>7385279

>>7385272
so you're saying it mistook "stairs" tag for something else due to token amount? hm, well its true there's way too many tokens in these so might be it

Anonymous

Anonymous Sat 20 May 2023 11:57:47 No.7385279 Report

Quoted By: >>7385291

>>7385275
that's the only reasonable explanation i can come up with, too much noise

Anonymous

Anonymous Sat 20 May 2023 12:06:23 No.7385284 Report

Quoted By: >>7385288 >>7385291 >>7385292

>>7385267
might be a dumb one, but are amount of images and their resolution matter? can they potentially contribute to errors im getting? i have 427 images and ~10 of them are scans of official art and they're higher than 4k res

Anonymous

Anonymous Sat 20 May 2023 12:07:59 No.7385288 Report

Quoted By: >>7385292

>>7385284
didnt somebody say last week they had issues with huge res images poisoning their dataset? try increasing bucket max res or resizing them smaller?

Anonymous

Anonymous Sat 20 May 2023 12:08:27 No.7385289 Report

Quoted By:

>>7384656
IDC for style, but I do consolidate for characters.

Anonymous

View Same Google ImgOps iqdb SauceNAO file.png, 159KiB, 254x866

Anonymous Sat 20 May 2023 12:08:47 No.7385291 Report

Quoted By: >>7385298 >>7385301 >>7385330

>>7385279
Right. Time to trim all the useless shit.
Would you happen to know how to tag images like this btw, i.e this skirt is a "suspender skirt" in every other image, but suspenders are not visible here, it still should be tagged as a suspender one, right?

>>7385284
99% sure thats a no. If anything it would be related related to the paging file being too small. if you've another GPU/integrated you can try disabling it in settings and then reinstaling torch as thats most likely whats causing it to not detect the right one

Anonymous

Anonymous Sat 20 May 2023 12:09:53 No.7385292 Report

Quoted By: >>7385301

>>7385288
hes talking about being unable to start training due to kernel errors with new torch

>>7385284
also how old is your GPU actually, maybe its simply incompatible with new torch due to being old?

Anonymous

View Same Google ImgOps iqdb SauceNAO dmg.png, 2MiB, 1024x1536

Anonymous Sat 20 May 2023 12:13:36 No.7385293 Report

Quoted By: >>7385295 >>7385296

im still using abyssorangemix2_hard , are there any similar but better models?

Anonymous

Anonymous Sat 20 May 2023 12:14:03 No.7385295 Report

Quoted By: >>7385297

>>7385293
yes

Anonymous

Anonymous Sat 20 May 2023 12:14:37 No.7385296 Report

Quoted By: >>7385297

>>7385293
cant tell if vaeless...

Anonymous

Anonymous Sat 20 May 2023 12:15:48 No.7385297 Report

Quoted By: >>7385299

>>7385295
insufferable
>>7385296
get your eyes checked

Anonymous

Anonymous Sat 20 May 2023 12:16:18 No.7385298 Report

Quoted By: >>7385311

>>7385291
first you need to determine if the model you're training on actually knows what a suspender skirt is, you could try interrogating the image in CLIP, just trying to generate it normally, or using a search engine like https://rom1504.github.io/clip-retrieval/?back=https%3A%2F%2Fknn.laion.ai&index=laion5B-H-14&useMclip=false

often times it is easier to not dial way in on specifying the type of clothing, you might get better results with just "skirt" and training the ai that "skirt" = the specific skirt your character has. once you've found out the best tags to train on the actual training should be easy

Anonymous

Anonymous Sat 20 May 2023 12:16:22 No.7385299 Report

Quoted By: >>7385303

>>7385297
>insufferable
I answered your question.

Anonymous

Anonymous Sat 20 May 2023 12:17:13 No.7385301 Report

Quoted By: >>7385311

>>7385291
passed through powershell in venv
torch.cuda.device_count() returns 1
torch.cuda.get_device_name(0) returns GTX1070

but if torch 1.12.1 won't work i'll hop onto BIOS and disable intergrated gpu

>>7385292
GTX1070 and i was training lora on torch 2.1 and even using AdamW8bit optimizer that supposed to throw errors on my series before reinstalling OS

Anonymous

Anonymous Sat 20 May 2023 12:18:23 No.7385303 Report

Quoted By: >>7385304 >>7385310

>>7385299
it wasn't my question. also doesn't change what i said. you are insufferable and this thread is better when you're away and people are posting actual images. of course, if you posted images, everyone would know exactly which shitty avatarfag you are.

Anonymous

Anonymous Sat 20 May 2023 12:19:09 No.7385304 Report

Quoted By: >>7385309

>>7385303
>are there

Anonymous

View Same Google ImgOps iqdb SauceNAO 00199-2398424177.png, 1MiB, 864x1536

Anonymous Sat 20 May 2023 12:19:24 No.7385306 Report

Quoted By:

Anonymous

Anonymous Sat 20 May 2023 12:20:42 No.7385309 Report

Quoted By: >>7385312

>>7385304
case in point. couldn't have made a better example if i tried

Anonymous

Anonymous Sat 20 May 2023 12:20:46 No.7385310 Report

Quoted By:

>>7385303
why nudging anons you don't like into wasting image slots bruh

Anonymous

Anonymous Sat 20 May 2023 12:21:33 No.7385311 Report

Quoted By: >>7385314

>>7385298
im training on base NAI, it should know. The issue is that would cause this model to generate only that one skirt of prompting "skirt" no? So specifying what type of skirt it is with the least tags possible and then removing general "skirt" should be the better route

>>7385301
>reinstalling OS
if you got all the dependencies you could possibly need including drivers then im unsure what else but integrated could be causing this.

Anonymous

Anonymous Sat 20 May 2023 12:22:08 No.7385312 Report

Quoted By: >>7385316

>>7385309
You feeling okay?

Anonymous

View Same Google ImgOps iqdb SauceNAO 00376-2273623944-nsfw, scan, (tr (...).png, 2MiB, 1024x1536

Anonymous Sat 20 May 2023 12:22:35 No.7385313 Report

Quoted By:

>>7384921
https://files.catbox.moe/0kkyna.png

Anonymous

Anonymous Sat 20 May 2023 12:23:01 No.7385314 Report

Quoted By: >>7385315

>>7385311
>should
there is no "should" with ai blackbox better test that shit

Anonymous

Anonymous Sat 20 May 2023 12:24:29 No.7385315 Report

Quoted By: >>7385327

>>7385314
I mean I did kinda test it before, when I've had issues with it overfitting on the same skirt no matter what I prompted, editing tags helped somewhat

Anonymous

Anonymous Sat 20 May 2023 12:25:12 No.7385316 Report

Quoted By: >>7385323

>>7385312
>953 posts
>265 files
it's sad niggers (you included) ruin these threads, but it is what it is

Anonymous

Anonymous Sat 20 May 2023 12:27:45 No.7385322 Report

Quoted By:

holy shit torch 1.12.1 finally trains my shit
>02:00<16:00:14
ITS SO FUCKING SLOW

Anonymous

Anonymous Sat 20 May 2023 12:28:13 No.7385323 Report

Quoted By: >>7385326

>>7385316
>he says not posting an image
Based.
Want to talk about your problems at home? I'm here for youm

Anonymous

Anonymous Sat 20 May 2023 12:29:09 No.7385326 Report

Quoted By: >>7385335

>>7385323
i am posting images. i'm also calling you a nigger. it's that simple, come up with better comebacks

Anonymous

Anonymous Sat 20 May 2023 12:29:39 No.7385327 Report

Quoted By: >>7385338

>>7385315
oh, i didn't realize you were overfitting, if you're getting that then you need more other stuff to balance it out

Anonymous

Anonymous Sat 20 May 2023 12:30:27 No.7385329 Report

Quoted By: >>7385334

im not posting images because im training

Anonymous

Anonymous Sat 20 May 2023 12:30:31 No.7385330 Report

Quoted By:

>>7385291
>Would you happen to know how to tag images like this btw, i.e this skirt is a "suspender skirt" in every other image, but suspenders are not visible here, it still should be tagged as a suspender one, right?
do not tag it as suspender skirt

Anonymous

View Same Google ImgOps iqdb SauceNAO what-are-u-doin.png, 705KiB, 1024x768

Anonymous Sat 20 May 2023 12:31:18 No.7385333 Report

Quoted By: >>7385336 >>7385342

<.<

Anonymous

Anonymous Sat 20 May 2023 12:31:37 No.7385334 Report

Quoted By:

>>7385329
no one wants your shitty style loras nor your nai tier gens that you think are soulful lol

Anonymous

Anonymous Sat 20 May 2023 12:32:41 No.7385335 Report

Quoted By: >>7385340

>>7385326
Tell me about your father.

Anonymous

View Same Google ImgOps iqdb SauceNAO 01797-3376960748.jpg, 120KiB, 1024x768

Anonymous Sat 20 May 2023 12:32:59 No.7385336 Report

Quoted By: >>7385342

>>7385333

Anonymous

Anonymous Sat 20 May 2023 12:33:16 No.7385338 Report

Quoted By:

>>7385327
I mean, isnt that always going to happen with oufits to some degree? There'll always be a "preference"

Anonymous

Anonymous Sat 20 May 2023 12:34:17 No.7385340 Report

Quoted By: >>7385345

>>7385335
>repeating same thing verbatim
zzzz

Anonymous

Anonymous Sat 20 May 2023 12:35:22 No.7385341 Report

Quoted By: >>7385348

is there any better alternative to dataset-tag-editor-standalone? if i load more than 20 images this gradio abomination starts lagging as hell

Anonymous

Anonymous Sat 20 May 2023 12:35:32 No.7385342 Report

Quoted By:

>>7385333
his dick broke

>>7385336
cute

Anonymous

Anonymous Sat 20 May 2023 12:36:14 No.7385345 Report

Quoted By: >>7385348

>>7385340
Come on, bud. I want to prevent a future mall shooting here.
Work with me.
Did he ever yell at you?

Anonymous

View Same Google ImgOps iqdb SauceNAO 00007-366441169.png, 2MiB, 1024x1536

Anonymous Sat 20 May 2023 12:40:48 No.7385348 Report

Quoted By: >>7385350 >>7385356 >>7385382

>>7385341
unfortunately no, but there's other issues if 20 images is causing that. i was just working with a 3k image dataset and it was doing fine
>>7385345
most leddit shit i've read itt

Anonymous

View Same Google ImgOps iqdb SauceNAO 01154-30346450.png, 2MiB, 1024x1536

Anonymous Sat 20 May 2023 12:41:16 No.7385349 Report

Quoted By: >>7385356 >>7385377

a month ago i followed a guide that said you should untag all clothing you want your character to have and i cant believe i listened to it

Anonymous

Anonymous Sat 20 May 2023 12:41:31 No.7385350 Report

Quoted By: >>7385358

>>7385348
>most leddit shit i've read itt
Know lots about leddit, do you? Might that be the root cause? Tell me more.

Anonymous

Anonymous Sat 20 May 2023 12:43:14 No.7385353 Report

Quoted By:

>>7384917
Catbox me that juicy pussy.

Anonymous

Anonymous Sat 20 May 2023 12:45:09 No.7385354 Report

Quoted By:

>>7384961
What's your regional prompter settings look like? I've been wanting to try but I can't figure out box placement or how to setup my prompt.

Anonymous

Anonymous Sat 20 May 2023 12:48:17 No.7385356 Report

Quoted By: >>7385357 >>7385358 >>7385360 >>7385364

>>7385348
might be a browser issue? i am using Brave and Chrome

>>7385349
is this guide https://rentry.org/ezlora#method-3-manual example of tagging any good? it says to untag character appearance and their clothes too

Anonymous

Anonymous Sat 20 May 2023 12:50:12 No.7385357 Report

Quoted By: >>7385361

>>7385356
>it says to untag character appearance and their clothes too
Kek

Anonymous

View Same Google ImgOps iqdb SauceNAO 00034-908602735.png, 3MiB, 1280x1920

Anonymous Sat 20 May 2023 12:50:13 No.7385358 Report

Quoted By: >>7385363

>>7385356
>might be a browser issue? i am using Brave and Chrome
doubt it, i'm on edge which is just chrome
>>7385350
you sure like lashing out at everyone with daddy issues. projection maybe?

Anonymous

View Same Google ImgOps iqdb SauceNAO 1654187637081162.jpg, 282KiB, 1408x1792

Anonymous Sat 20 May 2023 12:51:18 No.7385359 Report

Quoted By:

Anonymous

Anonymous Sat 20 May 2023 12:52:13 No.7385360 Report

Quoted By: >>7385377

>>7385356
>guide
In a way. Burning in character features can help, but its worse as it makes it hard to change them.
Its more useful for outfits where you consolidate an unique outfit to a single tag rather than 10 separate tags, but it suffers from the same issue of making it hard to change clothing details

Anonymous

Anonymous Sat 20 May 2023 12:52:59 No.7385361 Report

Quoted By: >>7385365

>>7385357
i mean if the goal is to be able to add the lora and have the character's default outfit show up without prompting the outfit or their appearance, then that is correct. it comes with other issues, though

Anonymous

View Same Google ImgOps iqdb SauceNAO 1682255311632542.jpg, 351KiB, 1408x1792

Anonymous Sat 20 May 2023 12:53:00 No.7385362 Report

Quoted By:

Anonymous

Anonymous Sat 20 May 2023 12:54:08 No.7385363 Report

Quoted By:

>>7385358
>everyone with daddy issues
Good, now that we have established that, we can get to the bottom of it.

Anonymous

Anonymous Sat 20 May 2023 12:55:04 No.7385364 Report

Quoted By: >>7385369 >>7385377

>>7385356
nah keep their clothes tagged, otherwise the train will consider all untagged components (like clothing) intrinsic to the train and basically glue them to your character
sometimes can't even remove them with (naked:1.5) because it doesn't recognize the clothes AS clothes
untag anything that would describe the actual, naked character (the poppy i posted, i removed tags like pointy ears, blue skin, purple eyes, twintails, etc) and consolidate clothing tags (ie: turn brown boots, black footwear, dark shoes into just boots)
there are probably more advanced methods of tagging but this works perfectly for me

Anonymous

Anonymous Sat 20 May 2023 12:55:10 No.7385365 Report

Quoted By: >>7385366

>>7385361
Imagine where we would be if not for that one retard who thought loras are embeddings, right?

Anonymous

Anonymous Sat 20 May 2023 12:56:48 No.7385366 Report

Quoted By: >>7385367

>>7385365
you do understand what training text enc does, no? or are you parroting shit you've heard other people say on here to fit in

Anonymous

Anonymous Sat 20 May 2023 12:58:49 No.7385367 Report

Quoted By: >>7385368

>>7385366
The thing about pruning (and activation tags) is that it was lifted straight out of embedding training and misapplied to loras. And then some Civit trainers decided that frying your clothes into your character is somehow desirable.
Kinda sad desu.

Anonymous

Anonymous Sat 20 May 2023 13:00:33 No.7385368 Report

Quoted By: >>7385370

>>7385367
i'll take that as a no, then

Anonymous

Anonymous Sat 20 May 2023 13:01:41 No.7385369 Report

Quoted By:

>>7385364
also including some images with wrong elements is usually fine (if you need to pad out the dataset) as long as it's tagged, my dataset had like 25% images of the character with white hair, but it doesn't generate white hair since i left that tag in
just don't have too many images with incorrect elements lol

Anonymous

Anonymous Sat 20 May 2023 13:02:45 No.7385370 Report

Quoted By:

>>7385368
Take it as what you want. It's true.

Anonymous

Anonymous Sat 20 May 2023 13:09:42 No.7385377 Report

Quoted By: >>7385380

>>7385360
>>7385364
in my first Kagami lora i tagged her appearance (twintails, hair bows, tsurime, ryouou school uniform) on all images, and to make her stop wearing default clothes i just added them into negatives while prompting other clothes (maid outfit, shrine maiden, bikini, nudity etc.) yet i haven't tested any alternative hairstyles

now i have much larger (427 instead of 25) dataset that have some alternative hairstyles like double bun and ponytail, and alternative clothes like shrine maiden, yet 80% of images does have her default look as described above, and i've tagged every image extensively just like first dataset. should i try untagging her default look and clothes or i'll run into this anon >>7385349 problems?

Anonymous

Anonymous Sat 20 May 2023 13:13:35 No.7385380 Report

Quoted By: >>7385391

>>7385377
if you have them all tagged properly and you're struggling to modify the clothes without adding those tags to the negatives, your lora is most likely overfit. the whole point of tagging all those things is so that when you don't prompt it, you don't get them

Anonymous

Anonymous Sat 20 May 2023 13:15:39 No.7385382 Report

Quoted By:

>>7385348
tatsuya lora

Anonymous

Anonymous Sat 20 May 2023 13:17:09 No.7385384 Report

Quoted By:

>>7384928
https://files.catbox.moe/itd7cm.png
https://files.catbox.moe/j2j0qg.png

Anonymous

Anonymous Sat 20 May 2023 13:17:26 No.7385385 Report

Quoted By:

>>7385101
inpaint guides are kinda shit

Anonymous

View Same Google ImgOps iqdb SauceNAO 03453-1601588487.jpg, 135KiB, 1024x768

Anonymous Sat 20 May 2023 13:26:19 No.7385390 Report

Quoted By:

Anonymous

View Same Google ImgOps iqdb SauceNAO 00007-142876200.png, 581KiB, 512x768

Anonymous Sat 20 May 2023 13:26:49 No.7385391 Report

Quoted By:

>>7385380
e10 has burned white and pink colorscheme into her clothes but it doesn't resemble her uniform, yet her twintails, hair bows and eye shape are still intact

pic related is (hiiragi kagami, 1girl, solo)

Anonymous

Anonymous Sat 20 May 2023 13:28:50 No.7385395 Report

Quoted By:

>>7385024
chatgpt ended up writing a working powershell script.
didn't have much duplicates it seems but man I'm not even sure if it's worth it to sort my loras. I'll try to add tags at least I guess

Anonymous

Anonymous Sat 20 May 2023 13:33:19 No.7385398 Report

Quoted By: >>7385406 >>7385522 >>7385654

>>7385101
NTA either, but the only things you need to concern yourself with are
Masked content: Fill is just noise; Original is well, original image plus whatever noise you have.
Inpaint area: Whole Picture takes the whole picture into account, so it's more integrated and uses your regular prompt; Only Masked makes the inpaint area like a whole new canvas and it doesn't take the rest of the image into account, so if you leave your prompt as is, it will regenerate an entire whole image inside that inpaint area. You need to specify only the thing you want to inpaint with Only Masked.
Denoise: How much of your inpaint area is changed. 0=no change, 100=completely new noise.
I do the MS Paint + inpaint routine (just slap a rough sketch of what you want onto your image and put it in inpaint), so I do Original, Whole Picture, and then denoise depending on the results I get, usually between 0.15-0.75. You can then integrate at low denoise more, or go straight to img2img highres at 0.15-0.5.
Only Masked can be useful in some circumstances, but I rarely need it. Oh, and you can obviously use topical loras as needed for each part of the picture.

Anonymous

Anonymous Sat 20 May 2023 13:47:55 No.7385406 Report

Quoted By: >>7385409

>>7385398
only masked is a lifesaver for poorshits like me. inpainting whole picture takes fucking years

Anonymous

Anonymous Sat 20 May 2023 13:50:17 No.7385409 Report

Quoted By: >>7385412

>>7385406
You have to lower the res to save time with that, right?
I dunno if you do that already, but I make a point to do most of the inpainting before upscaling.

Anonymous

Anonymous Sat 20 May 2023 13:52:32 No.7385412 Report

Quoted By:

>>7385409
>You have to lower the res to save time with that, right?
Yeah, I inpaint face/hands most of the time. You have to use a different prompt that describes what you are inpainting and use denoise below 0.5. Above 0.5 might work but it's a gacha.
I pretty much always set 512x512 resolution for only masked, works pretty well

Anonymous

View Same Google ImgOps iqdb SauceNAO file.png, 338KiB, 1523x789

Anonymous Sat 20 May 2023 14:08:41 No.7385423 Report

Quoted By:

>done all tag clearing
>still like 200+ tokens on some
I guess I'll just accept it, its not like the lora wasnt working before. so many weekend hours down the drain

Anonymous

Anonymous Sat 20 May 2023 14:27:23 No.7385445 Report

Quoted By:

It's kinda funny how AI absolutely ruined normalfags.

Anonymous

View Same Google ImgOps iqdb SauceNAO 01191-1090961316.png, 1MiB, 1024x1536

Anonymous Sat 20 May 2023 14:28:27 No.7385446 Report

Quoted By: >>7385453

cool hammer

Anonymous

Anonymous Sat 20 May 2023 14:30:13 No.7385447 Report

Quoted By: >>7385449 >>7385450 >>7385452 >>7385786

Anyway to hide the eyes? The AI really loves drawing eyes it seems

Anonymous

Anonymous Sat 20 May 2023 14:32:34 No.7385449 Report

Quoted By:

>>7385447
>hidden eyes

Anonymous

Anonymous Sat 20 May 2023 14:32:39 No.7385450 Report

Quoted By:

>>7385447
Negs or inpaint.

Anonymous

View Same Google ImgOps iqdb SauceNAO 00384-3084634624-best quality, o (...).png, 2MiB, 1024x1536

Anonymous Sat 20 May 2023 14:33:35 No.7385451 Report

Quoted By: >>7385458

baka

Anonymous

Anonymous Sat 20 May 2023 14:33:35 No.7385452 Report

Quoted By:

>>7385447
>blindfold

Anonymous

Anonymous Sat 20 May 2023 14:33:41 No.7385453 Report

Quoted By:

>>7385446
she took the sword and sandals 3 approach to combat

Anonymous

Anonymous Sat 20 May 2023 14:39:46 No.7385458 Report

Quoted By:

>>7385451
that expression is just inviting another anal session

Anonymous

View Same Google ImgOps iqdb SauceNAO Kiryuin Satzuri.jpg, 211KiB, 1024x1536

Anonymous Sat 20 May 2023 14:42:50 No.7385461 Report

Quoted By: >>7385474 >>7385484 >>7385512

Anonymous

Anonymous Sat 20 May 2023 14:49:07 No.7385474 Report

Quoted By:

>>7385461
based

Anonymous

View Same Google ImgOps iqdb SauceNAO catbox_9r2q44.png, 2MiB, 1024x1280

Anonymous Sat 20 May 2023 14:49:23 No.7385476 Report

Quoted By: >>7385482 >>7385921

Anonymous

View Same Google ImgOps iqdb SauceNAO 00092-1540700030.png, 1MiB, 960x1440

Anonymous Sat 20 May 2023 14:49:40 No.7385478 Report

Quoted By: >>7385482

Anonymous

Anonymous Sat 20 May 2023 14:53:13 No.7385482 Report

Quoted By:

>>7385476
goofy ah cock
>>7385478
based

Anonymous

Anonymous Sat 20 May 2023 14:53:35 No.7385484 Report

Quoted By: >>7385490

>>7385461
its missing something to hold her breasts, maybe the hand shoulda been there and her head in some kinda device.
Would jerk off to it but that part bothers me too much..

Anonymous

Anonymous Sat 20 May 2023 14:56:50 No.7385487 Report

Quoted By:

>>7384548
Catbox please

Anonymous

Anonymous Sat 20 May 2023 14:58:49 No.7385490 Report

Quoted By: >>7385496

>>7385484
what sort of device we talking
I don't do requests but I did have the same thoughts that something was missing

Anonymous

Anonymous Sat 20 May 2023 15:03:18 No.7385496 Report

Quoted By: >>7385512

>>7385490
Just something to restrain her head instead of the hand so she can still look like she's trying to free herself

Anonymous

Anonymous Sat 20 May 2023 15:04:52 No.7385499 Report

Quoted By: >>7385505

is there any new mixes for anime styles?
so far ive been forced to gen the images in 7th anime c to get the anime style and then img2img in abyessorangenswf to fix the overly thick lines.

Anonymous

Anonymous Sat 20 May 2023 15:08:32 No.7385505 Report

Quoted By: >>7385528

>>7385499
a few hundred probably

Anonymous

View Same Google ImgOps iqdb SauceNAO 00974-1345996553-best quality, ( (...).png, 618KiB, 656x960

Anonymous Sat 20 May 2023 15:11:03 No.7385509 Report

Quoted By:

Anonymous

Anonymous Sat 20 May 2023 15:13:24 No.7385512 Report

Quoted By: >>7385521

>>7385461
>>7385496
Reminds me of:
https://archiveofsins.com/h/thread/7276313/#7276392

Anonymous

View Same Google ImgOps iqdb SauceNAO 00385-1371881858-best quality, p (...).png, 2MiB, 1024x1536

Anonymous Sat 20 May 2023 15:14:11 No.7385514 Report

Quoted By:

Anonymous

Anonymous Sat 20 May 2023 15:17:52 No.7385518 Report

Quoted By:

unstable diffusion model training section is so fucking funny, its hard to believe such retarded people exist. Would recommend lurking as its hard to find anything this hilarious

Anonymous

Anonymous Sat 20 May 2023 15:18:48 No.7385520 Report

Quoted By:

you know what i find weird,
I can get something close to an art style just be training the textencoder.

Anonymous

Anonymous Sat 20 May 2023 15:19:02 No.7385521 Report

Quoted By:

>>7385512
lil too much but yeah

Anonymous

Anonymous Sat 20 May 2023 15:19:16 No.7385522 Report

Quoted By: >>7385527

>>7385398
NTA, but thanks for sharing your workflow. Do samplers for inpainting matter at all or is it the same Euler a, DPM ++SDE?

Anonymous

Anonymous Sat 20 May 2023 15:20:34 No.7385527 Report

Quoted By:

>>7385522
I don't think they do. At least not in my experience.

Anonymous

Anonymous Sat 20 May 2023 15:20:37 No.7385528 Report

Quoted By: >>7385529 >>7385532 >>7385561

>>7385505
i mean good ones,
i dont think it resonable to just run an xy plot on a hundred models.

Anonymous

Anonymous Sat 20 May 2023 15:21:07 No.7385529 Report

Quoted By:

>>7385528
to be more specific which of those models gets the anime style but doesn't have problems with thick lines?

Anonymous

Anonymous Sat 20 May 2023 15:21:49 No.7385532 Report

Quoted By:

>>7385528
probably tens of them around or you can merge yourself.

Anonymous

Anonymous Sat 20 May 2023 15:30:02 No.7385555 Report

Quoted By: >>7385561 >>7385563 >>7385566 >>7385571

>so many months have passed
>people still using NAI frankenstein finetunes
>progress has completely stagnated
Aiartistbros......what happened..

Anonymous

Anonymous Sat 20 May 2023 15:32:42 No.7385561 Report

Quoted By: >>7385564

>>7385528
It's not that bad if you exclude anything from civit.
>>7385555
Appreciate you're not genning 384x384px images vaguely resembling a face like you would be a year ago.

Anonymous

Anonymous Sat 20 May 2023 15:33:50 No.7385563 Report

Quoted By: >>7385569 >>7385573 >>7385579 >>7385580 >>7385944

>>7385555
Beyond complex perspectives, which are manageable with controlnet or img2img, is there anything that can't be made?
Any new concept can be rolled with a lora at will, people are just bottlenecked by imagination at this point.
Significant new development is pending VRAM/processing power barriers, or more reasonably, optimisation.

Anonymous

Anonymous Sat 20 May 2023 15:33:50 No.7385564 Report

Quoted By: >>7385573

>>7385561
>It's not that bad if you exclude anything from civit.
where else do people post mixes?

Anonymous

Anonymous Sat 20 May 2023 15:33:55 No.7385566 Report

Quoted By:

>>7385555
wow another doompost exactly like every other doompost how boring

Anonymous

Anonymous Sat 20 May 2023 15:37:34 No.7385569 Report

Quoted By:

>>7385563
>is there anything that can't be made?
there is one thing i havn't been able to do with all tools we currently have

Anonymous

Anonymous Sat 20 May 2023 15:38:35 No.7385571 Report

Quoted By:

>>7385555
if you're not being paid for those copypasterino doomposts i have bad news for you

Anonymous

Anonymous Sat 20 May 2023 15:39:55 No.7385573 Report

Quoted By:

>>7385563
To be fair, the models are technologically terrible and no progress on them is coming.
But it's still amazing what we can do because of a few leaks, so I'm not gonna be doomin.
>>7385564
HF is good.
99.9% of civit mixes are trash, and the 0.01% is copied from other places.

Anonymous

Anonymous Sat 20 May 2023 15:44:55 No.7385579 Report

Quoted By: >>7385589 >>7385596 >>7385937

>>7385563
Stuff like this is why /trash/ sdg is laughing at us we still using controlnet and LoRAs while furfags models have popular artists/characters/concepts integrated into their models

Anonymous

Anonymous Sat 20 May 2023 15:45:41 No.7385580 Report

Quoted By:

>>7385563
>people are just bottlenecked by imagination at this point.
thats a pretty big bottleneck for 99% of users to be fair

Anonymous

Anonymous Sat 20 May 2023 15:48:13 No.7385589 Report

Quoted By: >>7385593

>>7385579
furfags have way more finances and autism than animefags

i wonder if they already baked ashraely model to produce thick slutty anthro eevees on demand

Anonymous

Anonymous Sat 20 May 2023 15:50:12 No.7385593 Report

Quoted By: >>7385595

>>7385589
Nah, they prefer thick slutty feral eevees on demand.

Anonymous

Anonymous Sat 20 May 2023 15:51:45 No.7385595 Report

Quoted By:

>>7385593
as i said, autism

Anonymous

Anonymous Sat 20 May 2023 15:52:02 No.7385596 Report

Quoted By: >>7385599

>>7385579
you've just described overfitting anon

Anonymous

Anonymous Sat 20 May 2023 15:53:29 No.7385599 Report

Quoted By: >>7385602 >>7385616

>>7385596
I honestly don't get this psyop he's running here.

Anonymous

Anonymous Sat 20 May 2023 15:55:47 No.7385602 Report

Quoted By: >>7385606 >>7385616

>>7385599
there is literally one lonely nigger that cries over needing to drag slider below 1.00 but refusing to show his magnificent totally-not-overfit loras, let alone training them

Anonymous

Anonymous Sat 20 May 2023 15:57:12 No.7385606 Report

Quoted By:

>>7385602
???

Anonymous

View Same Google ImgOps iqdb SauceNAO furfags vs weebs.png, 195KiB, 1857x880

Anonymous Sat 20 May 2023 16:03:48 No.7385616 Report

Quoted By: >>7385621 >>7385627 >>7385632

>>7385599
>>7385602
NTA, I'm not making shit up /trash/ is mocking us

Anonymous

Anonymous Sat 20 May 2023 16:05:39 No.7385621 Report

Quoted By: >>7385626 >>7385642

>>7385616
they're furries anon, their very existence is a mockery.
Why would you give a shit, the last time their standalone model was posted it was laughed out of this place

Anonymous

Anonymous Sat 20 May 2023 16:08:00 No.7385626 Report

Quoted By: >>7385636

>>7385621
anon nobody actually looked at it, one guy just said it was shit and everyone took it as fact despite several other anons admitting that their thread & generation quality was miles better

Anonymous

Anonymous Sat 20 May 2023 16:08:01 No.7385627 Report

Quoted By:

>>7385616
>merely 30 minutes of inpainting
kek

Anonymous

Anonymous Sat 20 May 2023 16:10:30 No.7385632 Report

Quoted By: >>7385641

>>7385616
>giving a shit about what furries think

Anonymous

View Same Google ImgOps iqdb SauceNAO 00213-3535855236.png, 2MiB, 1024x1536

Anonymous Sat 20 May 2023 16:11:11 No.7385634 Report

Quoted By:

Anonymous

Anonymous Sat 20 May 2023 16:11:18 No.7385635 Report

Quoted By: >>7385637

We lost to https://files.catbox.moe/uo2mot.png
It's so over animebros...

Anonymous

Anonymous Sat 20 May 2023 16:11:37 No.7385636 Report

Quoted By: >>7385648

>>7385626
you made me look, their thread's far worse than ours in terms of images, what the fuck are you on about. go back

Anonymous

Anonymous Sat 20 May 2023 16:11:43 No.7385637 Report

Quoted By:

>>7385635
machoke on deez nuts

Anonymous

Anonymous Sat 20 May 2023 16:13:06 No.7385641 Report

Quoted By:

>>7385632
So your fine with furries being the leading face of AI? What's next bending over to them?

Anonymous

Anonymous Sat 20 May 2023 16:13:11 No.7385642 Report

Quoted By: >>7385644

>>7385621
we are weebs, our very existence is a mockery.
if they can make curvy large booba anthro pokemons with thick thighs they'll be clearly winning

Anonymous

Anonymous Sat 20 May 2023 16:14:40 No.7385644 Report

Quoted By: >>7385656

>>7385642
stop projecting schizoid. anime is mainstream in case you live under a rock and there was never anything wrong with it in the first place.
Go back to shitting in your diaper while fantasizing about dogs

Anonymous

Anonymous Sat 20 May 2023 16:15:42 No.7385648 Report

Quoted By: >>7385651

>>7385636
idc about their thread im just saying ive seen this sentiment several times by now

Anonymous

Anonymous Sat 20 May 2023 16:16:55 No.7385651 Report

Quoted By:

>>7385648
it was doomposting bubby

Anonymous

Anonymous Sat 20 May 2023 16:18:17 No.7385654 Report

Quoted By: >>7385661

>>7385398
thanks man, I'll screencap this

just a quick question, what sampler do you use for inpaint and would most of this also apply to controlnet since I'd like to inpaint stuff using depth maps of hands etc. ?

Anonymous

Anonymous Sat 20 May 2023 16:20:10 No.7385656 Report

Quoted By: >>7385667

>>7385644
you missed whole point when you've changed "weeb" to "anime"

Anonymous

Anonymous Sat 20 May 2023 16:20:56 No.7385661 Report

Quoted By: >>7385665

>>7385654
I use the same that I'm genning with. Euler A, was using SDE Karras before. Only tidbit I have is that DDIM doesn't work for upscaling.
I dunno about controlnet, I haven't been using it.

Anonymous

Anonymous Sat 20 May 2023 16:23:23 No.7385665 Report

Quoted By:

>>7385661
cheers, thanks for explaining the stuff before

Anonymous

Anonymous Sat 20 May 2023 16:24:10 No.7385667 Report

Quoted By:

>>7385656
weeb is a buzzword, zoomer

Anonymous

Anonymous Sat 20 May 2023 16:27:38 No.7385670 Report

Quoted By: >>7385672 >>7385760

So with controlnet reference thingy you can turn old ps1 screenshots into detailed images or was that a meme

Anonymous

Anonymous Sat 20 May 2023 16:28:41 No.7385672 Report

Quoted By: >>7385760 >>7387262

>>7385670
>he doesn't want to turn detailed hentai images into old ps1 screenshots
ngmi

Anonymous

View Same Google ImgOps iqdb SauceNAO 00002-2376426310-nsfw, scan, (tr (...).png, 2MiB, 1024x1536

Anonymous Sat 20 May 2023 16:30:17 No.7385676 Report

Quoted By: >>7385760

Anonymous

Anonymous Sat 20 May 2023 16:48:57 No.7385704 Report

Quoted By: >>7385760

is the AI booru gone?

Anonymous

Anonymous Sat 20 May 2023 17:02:23 No.7385728 Report

Quoted By:

owari da...

Anonymous

View Same Google ImgOps iqdb SauceNAO cf5c17b57f0e25f53c2b21b9b584ed0c (...).png, 1MiB, 1024x1536

Anonymous Sat 20 May 2023 17:08:18 No.7385734 Report

Quoted By: >>7385740

that's a lotta cutouts

Anonymous

View Same Google ImgOps iqdb SauceNAO 00048-3597351280.png, 519KiB, 512x768

Anonymous Sat 20 May 2023 17:10:12 No.7385737 Report

Quoted By: >>7385746 >>7385787

>>7385144
I was merely standing on giants

Anonymous

View Same Google ImgOps iqdb SauceNAO 00192-1747886457-nsfw, scan, (tr (...).png, 2MiB, 1024x1536

Anonymous Sat 20 May 2023 17:13:06 No.7385740 Report

Quoted By: >>7385928

>>7385734
It happens but at least it works.

Anonymous

Anonymous Sat 20 May 2023 17:16:47 No.7385746 Report

Quoted By:

>>7385737
The cutest...

Anonymous

Anonymous Sat 20 May 2023 17:18:00 No.7385749 Report

Quoted By: >>7385773

>furry psyop again
sorry I will never become a furfag

Anonymous

Anonymous Sat 20 May 2023 17:20:17 No.7385757 Report

Quoted By:

The furry models are kinda impressive at IP recognition.
The quality is pretty shit tho.

Anonymous

View Same Google ImgOps iqdb SauceNAO 00030-539310393.png, 339KiB, 480x640

Anonymous Sat 20 May 2023 17:20:46 No.7385760 Report

Quoted By: >>7387262

why isn't controlnet processed result stored in png info?

>>7385670
you can but >>7385672

>>7385676
based

>>7385704
yeah, rip

Anonymous

Anonymous Sat 20 May 2023 17:23:17 No.7385765 Report

Quoted By: >>7385768 >>7385771

why are nipples getting fucked up so often?

Anonymous

Anonymous Sat 20 May 2023 17:25:47 No.7385768 Report

Quoted By:

>>7385765
They are?

Anonymous

Anonymous Sat 20 May 2023 17:27:17 No.7385771 Report

Quoted By: >>7385894

>>7385765
this is just my own shizo observations, but loras based on artists who don't draw full nudity tend to have fucked up nipples

Anonymous

Anonymous Sat 20 May 2023 17:34:58 No.7385773 Report

Quoted By: >>7385774 >>7385775

>>7385749
i've skimmed through archived /trash/ threads and it feels like scrolling devianart, good bits here and there like this https://files.catbox.moe/u66yqo.png that could bridge weeb into furry stuff
but the rest is either photorealistic besiality, close-up anuses of male dragons with (extremely detailed:1.4) slapped on their scales or someone stuffing horse dildos into extremely buffed and blurry male pokemons because noone tells anon to use proper upscaler

seems like they don't have sameface disease like NAI mixes does, though

Anonymous

Anonymous Sat 20 May 2023 17:36:14 No.7385774 Report

Quoted By:

>>7385773
>close-up anuses of male dragons with (extremely detailed:1.4)
BASED

Anonymous

Anonymous Sat 20 May 2023 17:36:15 No.7385775 Report

Quoted By:

>>7385773
inb4
>fuck off furry psyop

Anonymous

View Same Google ImgOps iqdb SauceNAO 00504-2906486222, DPM++ 2M Karra (...).jpg, 659KiB, 768x1280

Anonymous Sat 20 May 2023 17:48:30 No.7385786 Report

Quoted By: >>7385793

>>7385447
my favorite is closed eyes

Anonymous

Anonymous Sat 20 May 2023 17:48:36 No.7385787 Report

Quoted By:

>>7385737
How do you get a result like that, jesus

Anonymous

Anonymous Sat 20 May 2023 17:50:51 No.7385792 Report

Quoted By: >>7385808 >>7385809

>>7384536
stop using python 3.9

Anonymous

Anonymous Sat 20 May 2023 17:51:29 No.7385793 Report

Quoted By: >>7385806

>>7385786
gatoboxu

Anonymous

View Same Google ImgOps iqdb SauceNAO 00569-1335680527, DPM++ 2M Karra (...).jpg, 547KiB, 768x1280

Anonymous Sat 20 May 2023 18:01:22 No.7385806 Report

Quoted By:

>>7385793
https://files.catbox.moe/ydv78f.jpg

Anonymous

Anonymous Sat 20 May 2023 18:01:56 No.7385808 Report

Quoted By: >>7385809

>>7385792
I'm too poor so tell that to Paperspace.

Anonymous

Anonymous Sat 20 May 2023 18:02:37 No.7385809 Report

Quoted By: >>7386002

>>7385792
>>7385808
lol I went to check the PRs and somebody actually fixed it just now. https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/10578

Anonymous

Anonymous Sat 20 May 2023 18:03:07 No.7385810 Report

Quoted By: >>7385811 >>7385813

any characters that have plenty of fanart that someone needs trained?

Anonymous

Anonymous Sat 20 May 2023 18:05:02 No.7385811 Report

Quoted By:

>>7385810
literally all of them

Anonymous

Anonymous Sat 20 May 2023 18:06:56 No.7385813 Report

Quoted By:

>>7385810
There's no good Junko Enoshima LoRA from what I've seen.

Anonymous

View Same Google ImgOps iqdb SauceNAO 00241-2624994497.png, 596KiB, 704x768

Anonymous Sat 20 May 2023 18:34:53 No.7385834 Report

Quoted By:

Anonymous

Anonymous Sat 20 May 2023 18:43:23 No.7385847 Report

Quoted By:

Genned this: >>7385846

Anonymous

View Same Google ImgOps iqdb SauceNAO 00013-3180097495.png, 732KiB, 768x1024

Anonymous Sat 20 May 2023 18:55:12 No.7385867 Report

Quoted By: >>7385914

Anonymous

View Same Google ImgOps iqdb SauceNAO demonmommy.png, 2MiB, 1433x1433

Anonymous Sat 20 May 2023 19:03:36 No.7385880 Report

Quoted By:

Something I made for a friend

Anonymous

Anonymous Sat 20 May 2023 19:09:37 No.7385894 Report

Quoted By:

>>7385771
interesting. i've been wondering this myself as a noob: is it better to just have full nude datasets with no clothed images?

Anonymous

View Same Google ImgOps iqdb SauceNAO 00780-3538506509.png, 4MiB, 1728x2592

Anonymous Sat 20 May 2023 19:11:52 No.7385901 Report

Quoted By: >>7385920

There's something more erotic about it not being nude.

Anonymous

Anonymous Sat 20 May 2023 19:22:35 No.7385914 Report

Quoted By:

>>7385867
gato?

Anonymous

Anonymous Sat 20 May 2023 19:26:53 No.7385920 Report

Quoted By: >>7386001

>>7385901
Teasefags always win.

Anonymous

Anonymous Sat 20 May 2023 19:27:48 No.7385921 Report

Quoted By:

>>7385476
Achromaru LoRA please

Anonymous

Anonymous Sat 20 May 2023 19:32:52 No.7385928 Report

Quoted By:

>>7385740
>2 navels

so close

Anonymous

View Same Google ImgOps iqdb SauceNAO 02821-2363338844.png, 1MiB, 1024x1536

Anonymous Sat 20 May 2023 19:34:59 No.7385930 Report

Quoted By:

the universal solvent mix is pretty coolio

Anonymous

Anonymous Sat 20 May 2023 19:38:10 No.7385933 Report

Quoted By: >>7386485

>>7385211
I just want regular bras man. The ones default when proompting aren't that good.

Anonymous

Anonymous Sat 20 May 2023 19:39:34 No.7385936 Report

Quoted By:

Is there a way to use higher batch size when making grids? That is to say without getting several images per x/y but instead one for every at the same time.

Anonymous

Anonymous Sat 20 May 2023 19:39:46 No.7385937 Report

Quoted By: >>7386488

>>7385579
Seriously can you FUCK OFF already? Stop shilling those hilariously garbage models for your trashy disgusting fetish you freak

Anonymous

View Same Google ImgOps iqdb SauceNAO 02845-245137060.png, 400KiB, 512x768

Anonymous Sat 20 May 2023 19:41:36 No.7385939 Report

Quoted By: >>7385972

>>7385177
bruh it shouldn't be too hard no? Panties does underwear perfectly, for example, and so does bra. No idea why you're struggling

Anonymous

Anonymous Sat 20 May 2023 19:43:46 No.7385944 Report

Quoted By:

>>7385563
I've been struggling with complex concepts such as "shota hugging tall girl from behind while rubbing his cock between her thighs"

and mind you, I've almost done it, but then it breaks in details such as the spine of the shota being too long compared to where his dick is

Anonymous

View Same Google ImgOps iqdb SauceNAO 00008-1015549546.png, 666KiB, 512x768

Anonymous Sat 20 May 2023 19:44:55 No.7385946 Report

Quoted By: >>7385950 >>7385967

God I love genning big titty hime cuts. Only issue is liking so many of them I'm forced to fill my favorites folder

Anonymous

Anonymous Sat 20 May 2023 19:45:44 No.7385950 Report

Quoted By: >>7385954

>>7385946
enter: bursting breasts tag and lora, and boobage tag

Anonymous

Anonymous Sat 20 May 2023 19:47:40 No.7385954 Report

Quoted By: >>7385961

>>7385950
Don't know what boobage tag is but you can be damn sure I'm using the bursting breasts lora

Anonymous

View Same Google ImgOps iqdb SauceNAO 00151-2907767031.png, 4MiB, 1728x2304

Anonymous Sat 20 May 2023 19:51:00 No.7385960 Report

Quoted By: >>7385964

Anonymous

Anonymous Sat 20 May 2023 19:51:32 No.7385961 Report

Quoted By:

>>7385954
it's tits bondage which is weird by itself, but, used right, it adds skindentation to skimpy tops and bursting breasts

Anonymous

Anonymous Sat 20 May 2023 19:52:33 No.7385964 Report

Quoted By: >>7385979 >>7386038

>>7385960
das noice, can you catbox me up captain?

any trick for achieving such a resolution? Is it hires fix or upscale?

Anonymous

Anonymous Sat 20 May 2023 19:53:08 No.7385967 Report

Quoted By: >>7385969 >>7385971

>>7385946
For some reason the hime cut tag never works for me, I'm not really sure if it's just a model issue

Anonymous

Anonymous Sat 20 May 2023 19:54:02 No.7385969 Report

Quoted By: >>7385978

>>7385967
Make sure to tag blunt bangs as well, works for me most of the time.

Anonymous

Anonymous Sat 20 May 2023 19:54:34 No.7385971 Report

Quoted By:

>>7385967
Same

Anonymous

Anonymous Sat 20 May 2023 19:54:39 No.7385972 Report

Quoted By:

>>7385939
It's that easy? Ill have to try later i guess.

Anonymous

Anonymous Sat 20 May 2023 19:57:55 No.7385978 Report

Quoted By:

>>7385969
Damn, you weren't kidding. Thanks, this changes everything

Anonymous

Anonymous Sat 20 May 2023 19:58:14 No.7385979 Report

Quoted By:

>>7385964
NTA but if you send your gen to extras it allowed you to upscale it further
only adds extra clarity if you zoom in really

Anonymous

Anonymous Sat 20 May 2023 20:09:48 No.7385991 Report

Quoted By: >>7385998 >>7386010

>got literally addicted to proompting
>I spend the entire day working from home dodging office work to proompt
>proompt during meetings
>fap so much I just stopped using tinder and meeting with real girls
>started getting blisters on my ass for sitting and prompting for 12h+ sessions
>realize that at this rate it could literally kill me
>get standing desk
>get gym equipment
>lift weights and do pull ups/push ups/leg raises/squats in between prompts
>after only a few months of this starts to actually see results

I-is it good or bad?

Anonymous

Anonymous Sat 20 May 2023 20:11:17 No.7385992 Report

Quoted By: >>7386004

its shit, delete it, saddest fake greentext I've seen in a long time, it never happened so hard I wouldnt even buy

Anonymous

Anonymous Sat 20 May 2023 20:15:05 No.7385998 Report

Quoted By:

>>7385991
>he thinks we think he ever saw a woman irl
lmao wrong place baitkun

Anonymous

View Same Google ImgOps iqdb SauceNAO 00917-697486347.png, 2MiB, 1536x1536

Anonymous Sat 20 May 2023 20:15:21 No.7386001 Report

Quoted By: >>7386029

>>7385920
Inspired

Anonymous

Anonymous Sat 20 May 2023 20:16:18 No.7386002 Report

Quoted By:

>>7385809
>immediately broken again after Voldy merged the UI selection for cross attention optimization branch
rip, sucks to be a poorperspacefag

Anonymous

Anonymous Sat 20 May 2023 20:16:34 No.7386004 Report

Quoted By:

>>7385992
>4channer goes in denial to the idea of physical activity being good for you

Anonymous

Anonymous Sat 20 May 2023 20:16:38 No.7386006 Report

Quoted By: >>7386014

>12 hours
>blisters
lmao not believable in the least. Although I can relate with the addicted to proompting bit, I was as well back in september/october and the start of NAI.

Anonymous

Anonymous Sat 20 May 2023 20:18:22 No.7386009 Report

Quoted By: >>7386012 >>7386016 >>7386019

How do you guys handle knees, elbows and belly buttons turning into nipples?

Anonymous

Anonymous Sat 20 May 2023 20:18:26 No.7386010 Report

Quoted By: >>7386018

>>7385991
sometimes, I prompt so hard, I rip the skin

Anonymous

Anonymous Sat 20 May 2023 20:18:46 No.7386012 Report

Quoted By: >>7386015

>>7386009
I rarely ever need to prompt nipples.

Anonymous

Anonymous Sat 20 May 2023 20:19:01 No.7386014 Report

Quoted By: >>7386024

>>7386006
I'm a skinny guy and my shitty ass ikea chair has a completely flattened cushion

Anonymous

Anonymous Sat 20 May 2023 20:19:26 No.7386015 Report

Quoted By:

>>7386012
It happens without me prompting them

Anonymous

Anonymous Sat 20 May 2023 20:20:03 No.7386016 Report

Quoted By:

>>7386009
inpaint

but I rarely have this problem, there must be something off in your proompting

Anonymous

Anonymous Sat 20 May 2023 20:20:11 No.7386018 Report

Quoted By:

>>7386010
my daddy taught us

Anonymous

Anonymous Sat 20 May 2023 20:20:13 No.7386019 Report

Quoted By:

>>7386009
dont remember the last time i thappened.
Probably fixed by loras

Anonymous

Anonymous Sat 20 May 2023 20:21:29 No.7386024 Report

Quoted By: >>7386028

>>7386014
I'm a skinny guy and my IKEA Markus has served me well for 8 years with no blisters.

Anonymous

View Same Google ImgOps iqdb SauceNAO 02861-3993622250.png, 2MiB, 1024x1536

Anonymous Sat 20 May 2023 20:21:42 No.7386025 Report

Quoted By:

>tfw no slutty curvy shortstack to ask you to empty your balls on her

Anonymous

Anonymous Sat 20 May 2023 20:23:27 No.7386028 Report

Quoted By:

>>7386024
the markus has garbage back support, but the seat is very sturdy yes, I got a mesh one that supports the back way better but has this flattening problem, my ass is basically on wood after 3-4 years of using it. I'm totally going for a herman miller tier chair next time, since I can afford it now

Anonymous

Anonymous Sat 20 May 2023 20:23:41 No.7386029 Report

Quoted By: >>7386035

>>7386001
Can I get a catbox? I've been trying to do something condoms but I can't get it to work for my life

Anonymous

Anonymous Sat 20 May 2023 20:24:44 No.7386032 Report

Quoted By:

>>7382726
>>7383628
for next bread remember to use updated ones

Anonymous

Anonymous Sat 20 May 2023 20:25:39 No.7386035 Report

Quoted By: >>7386054

>>7386029
https://litter.catbox.moe/e86399.png

Anonymous

Anonymous Sat 20 May 2023 20:26:54 No.7386038 Report

Quoted By: >>7386040

>>7385964
https://files.catbox.moe/exn0vt.png
This is the original image prior to inpainting to get rid of Vex's weird shadow guy.

Honestly my graphics card is shit (it's on a laptop), so I gen with Hires fix until i get something decent, slap it into inpainting to do touch-ups, then run it through the SD Ultimate Upscaler script to increase resolution.

Anonymous

Anonymous Sat 20 May 2023 20:28:31 No.7386040 Report

Quoted By: >>7386044 >>7386495

>>7386038
>sd ultimate upscaler

Mmmh I should really give it a go, but every tinme I tried it, it'd split the prompt into squares instead of blending them together

Anonymous

Anonymous Sat 20 May 2023 20:31:41 No.7386044 Report

Quoted By: >>7386047

>>7386040
set denoising strength to 0.15-0.2 and you should be good. I also removing the prompt beside the loras which helps(?)

Anonymous

Anonymous Sat 20 May 2023 20:33:39 No.7386047 Report

Quoted By:

>>7386044
thanks, I'll try!

Anonymous

Anonymous Sat 20 May 2023 20:35:47 No.7386049 Report

Quoted By: >>7386051 >>7386055 >>7386064 >>7386068

will never not be funny how they deny inpaint anon's masterpieces but allow vaeless shit https://aibooru.online/posts/30771

Anonymous

Anonymous Sat 20 May 2023 20:36:38 No.7386051 Report

Quoted By: >>7386053 >>7386068

>>7386049
>they deny inpaint anon's masterpieces
They do?

Anonymous

Anonymous Sat 20 May 2023 20:37:03 No.7386053 Report

Quoted By: >>7386068

>>7386051
Pretty sure he complained about it a few times?

Anonymous

Anonymous Sat 20 May 2023 20:38:47 No.7386054 Report

Quoted By:

>>7386035
Thanks anon, would you have any idea for how to have a decent looking condom on a dick? There's a lora on civitai but I've been struggling with it for the past 2 days because it seems like its absolute shit

Anonymous

Anonymous Sat 20 May 2023 20:39:50 No.7386055 Report

Quoted By: >>7386056 >>7386057 >>7386069 >>7386501

>>7386049
There should be some general quality minimum for AI boorus seeing as people can just shit out thousands of these shitty images every hour. Other boorus are at least limited by the effort it takes to produce an image.
Also does anyone browse AI boorus? I've never done it myself

Anonymous

Anonymous Sat 20 May 2023 20:41:38 No.7386056 Report

Quoted By:

>>7386055
>quality minimum
subjective. Even if you say they should deny limb fuckups etc you'd have to set a comprehensive standard which can be rough.
>does anyone
someone must be. I used to check if someone made art of niche charas but no one does so its w/e

Anonymous

Anonymous Sat 20 May 2023 20:42:04 No.7386057 Report

Quoted By:

>>7386055
moderating a booru made for images that can be shat out in mere second and still look somewhat decent would take up your entire day nobodys gonna want to do it

Anonymous

Anonymous Sat 20 May 2023 20:42:06 No.7386058 Report

Quoted By: >>7386059

new bread?

Anonymous

Anonymous Sat 20 May 2023 20:44:03 No.7386059 Report

Quoted By:

>>7386058
oven broken

Anonymous

Anonymous Sat 20 May 2023 20:50:46 No.7386064 Report

Quoted By: >>7386072

>>7386049
https://cdn.aibooru.online/sample/92/18/sample-92189b6dc94ebfa98da793af3e618a82.jpg

as if this isn't inpainted as fuck, for example

Anonymous

Anonymous Sat 20 May 2023 20:52:26 No.7386068 Report

Quoted By: >>7386074 >>7386077 >>7386507

>>7386049
>>7386051
>>7386053
I've advocated for moving off of the booru if your goal is to share content. Pixiv is just far more convenient. You upload something, tag it if you want, add some flags like R18/AI and you're pretty much done.
AIBooru requires so much more effort in regards to tagging and your stuff, well, some of my stuff got trashed anyway.
It's still on the site, just filtered. I just can't be bothered going through all the extra work for a gamble on whatever the reviewers feel. Also they filtered the 2B images on the grounds of non-con which was a shame. She was built to be (ab)used why is this controversial.
Some stuff on AIBooru, most stuff on Pixiv or /hdg/.

Anonymous

Anonymous Sat 20 May 2023 20:53:02 No.7386069 Report

Quoted By:

>>7386055
Impossible to curate. The volume is too massive. Who's going to do that? And for free too.

Anonymous

Anonymous Sat 20 May 2023 20:53:38 No.7386072 Report

Quoted By: >>7386079

>>7386064
wdym? that it shouldnt be allowed because its inpainted?

Anonymous

Anonymous Sat 20 May 2023 20:53:59 No.7386074 Report

Quoted By: >>7386083

>>7386068
does pixiv allow questionable content or did they also purge stuff like rapey scenarios and flatties?

Anonymous

Anonymous Sat 20 May 2023 20:54:39 No.7386077 Report

Quoted By:

>>7386068
>NON-CON
IBHDHIUSHDGFIBSFSAF
AAHHAHAHAHA
ahahsahsafhiusahiuafh
ahahahahhaahhhahahahhahahahahahahahahahahahahahahaha
>NON CON BANNED ON A BOORU
HAHAHAHAHAHAHAHAHAHA

Anonymous

Anonymous Sat 20 May 2023 20:55:00 No.7386079 Report

Quoted By: >>7386083

>>7386072
that's not on me, it's on aibooru rules it seems, they don't allow heavily reworked ai gens

Anonymous

Anonymous Sat 20 May 2023 20:56:36 No.7386083 Report

Quoted By: >>7386092

>>7386079
lmfao
I fucking cant
what a fucking joke of a site AHAHAHAHAHA
you gotta post vaeless to get approved bro, if ur better than the mods ur not getting in

>>7386074
pixiv allows pretty much everything that isnt photorealistic CP

Anonymous

Anonymous Sat 20 May 2023 20:58:24 No.7386086 Report

Quoted By: >>7386091

/hdg/ is over...

Anonymous

Anonymous Sat 20 May 2023 20:59:49 No.7386089 Report

Quoted By: >>7386098

need a new bread lads

Anonymous

Anonymous Sat 20 May 2023 21:00:08 No.7386091 Report

Quoted By:

>>7386086
again...

Anonymous

Anonymous Sat 20 May 2023 21:00:10 No.7386092 Report

Quoted By: >>7386100

>>7386083
including english language stuff? I may be creating an account for my cringey captions there, to get some suggestions and feedbacks

Anonymous

Anonymous Sat 20 May 2023 21:04:35 No.7386098 Report

Quoted By:

>>7386089
non
i refuse

Anonymous

Anonymous Sat 20 May 2023 21:06:00 No.7386100 Report

Quoted By: >>7386104

>>7386092
wdym eng language stuff

Anonymous

Anonymous Sat 20 May 2023 21:07:18 No.7386104 Report

Quoted By:

>>7386100
If the board wasn't broken I'd show you, I'm that guy that prompts scenarios and adds speech balloons

Anonymous

Anonymous Sat 20 May 2023 21:07:20 No.7386106 Report

Quoted By: >>7386112 >>7386113 >>7386115 >>7386127

bread

>>7386102
>>7386102
>>7386102

Anonymous

Anonymous Sat 20 May 2023 21:07:26 No.7386107 Report

Quoted By: >>7386117

remember to use updated OP and copy links to other boards from old OP aswell

Anonymous

Anonymous Sat 20 May 2023 21:07:31 No.7386108 Report

Quoted By: >>7386112 >>7386113 >>7386115 >>7386127

>>7386103

Anonymous

Anonymous Sat 20 May 2023 21:08:19 No.7386112 Report

Quoted By: >>7386117

>>7386106
>>7386108
which one?

Anonymous

Anonymous Sat 20 May 2023 21:08:27 No.7386113 Report

Quoted By: >>7386124

>>7386106
>>7386108
are you two retarded, 2 people have said to use updated one by now and you still cant fucking copy it? Delete and start over

Anonymous

Anonymous Sat 20 May 2023 21:08:39 No.7386115 Report

Quoted By:

>>7386106
>>7386108
>1 post apart
kek...

Anonymous

Anonymous Sat 20 May 2023 21:08:50 No.7386117 Report

Quoted By:

>>7386112
Well shit. Apparently neither because of >>7386107. Someone else bake.

Anonymous

Anonymous Sat 20 May 2023 21:10:42 No.7386124 Report

Quoted By:

>>7386113
I blame my regex filters for being too strong

Anonymous

Anonymous Sat 20 May 2023 21:11:37 No.7386127 Report

Quoted By: >>7386136

>>7386106
but I liked the Reimu op pic, rebake, this time with updated resources >>7386108
forgot the logo anyway

Anonymous

Anonymous Sat 20 May 2023 21:13:06 No.7386135 Report

Quoted By: >>7386138

use

X Edition

Old: XXX

>WEBUI SETUP
Nvidia: https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Install-and-Run-on-NVidia-GPUs
AMD: https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Install-and-Run-on-AMD-GPUs
Cloud: https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Online-Services
Optimizations: https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Optimizations
Extensions: https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Extensions
Custom Scripts: https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Custom-Scripts

>RESOURCES
Wiki: https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki
LoRAs/models: https://civitai.com/ | https://huggingface.co/ | https://gitgud.io/gayshit/makesomefuckingporn
Training: https://github.com/derrian-distro/LoRA_Easy_Training_Scripts | https://rentry.org/59xed3
Tags: https://danbooru.donmai.us/wiki_pages/tag_groups
ControlNet: https://rentry.org/dummycontrolnet
LamaCleaner: https://huggingface.co/spaces/Sanster/Lama-Cleaner-lama
Animation: https://rentry.org/AnimAnon
Wildcards: https://rentry.org/NAIwildcards
4chanX Catbox userscript: [Gist] catbox-README.md
OP Template/Logos: https://pastebin.com/XkPuPQje | https://files.catbox.moe/xk99gy.png | https://files.catbox.moe/35bdt5.gif

>Related
>>>/aco/sdg
>>>/g/sdg
>>>/d/ddg
>>>/e/sdg
>>>/vt/vtai
>>>/vg/aids

Anonymous

Anonymous Sat 20 May 2023 21:13:07 No.7386136 Report

Quoted By: >>7386138

>>7386127
Quote the exact post so I don't mess it up next time. Like I said, my filters are probably too strong and I missed it.

Anonymous

Anonymous Sat 20 May 2023 21:14:07 No.7386138 Report

Quoted By: >>7386146 >>7386168

>>7386136
see >>7386135
might wanna change order of lora repos so ours is first but up to you

Anonymous

Anonymous Sat 20 May 2023 21:16:31 No.7386146 Report

Quoted By: >>7386168

>>7386138
Ours should definitely be first I think. The setup links are a bit verbose and could be limited to 1 just for the wiki but it does help with discoverability. Also my catbox script link is broken in that.
Just migrate to >>7386103 and hopefully I'll be around to bake next time.

Anonymous

Anonymous Sat 20 May 2023 21:24:17 No.7386168 Report

Quoted By:

>>7386138
>>7386146
I like it at the end because you just highlight from the back and drag in. If it's in the middle I gotta try and be precise.

Anonymous

Anonymous Sat 20 May 2023 22:08:30 No.7386272 Report

Quoted By:

>>7384787
catbox please

Anonymous

Anonymous Sat 20 May 2023 23:54:11 No.7386485 Report

Quoted By:

>>7385933
they're either dumb or purposefully misunderstanding you. i second the request for a lora that specializes in a variety of highly detailed underwear

Anonymous

Anonymous Sat 20 May 2023 23:54:49 No.7386488 Report

Quoted By:

>>7385937
who hurt you

Anonymous

Anonymous Sat 20 May 2023 23:57:50 No.7386495 Report

Quoted By:

>>7386040
use either tiled diffusion or controlnet tile sampler + tile model along side ultimate SD

Anonymous

Anonymous Sun 21 May 2023 00:00:58 No.7386501 Report

Quoted By:

>>7386055
having a view that groups search results by author and putting a star rating on that authors preview image for whatever tag you searched would help tremendously. then add a mouse hover preview to view more photos by that author without needing to navigate to a new page would be top

Anonymous

Anonymous Sun 21 May 2023 00:04:47 No.7386507 Report

Quoted By:

>>7386068
moralfags have no business modding a booru. unless it's illegal, you can kindly stfu with your thoughts on xyz kink

Anonymous

Anonymous Sun 21 May 2023 04:09:23 No.7386822 Report

Quoted By:

>>7385084
its on gayshit
https://rentry.org/shitty_loras

Anonymous

Anonymous Sun 21 May 2023 08:50:50 No.7387202 Report

Quoted By: >>7387208

>>7384875
what model is that?

Anonymous

Anonymous Sun 21 May 2023 08:58:26 No.7387208 Report

Quoted By: >>7387210

>>7387202
Model hash: d3cd6ac55a, Model: hassakuHentaiModel_v12

Anonymous

Anonymous Sun 21 May 2023 09:00:33 No.7387210 Report

Quoted By: >>7387216

>>7387208
Where get?

Anonymous

Anonymous Sun 21 May 2023 09:05:34 No.7387216 Report

Quoted By:

>>7387210
civitai

Anonymous

Anonymous Sun 21 May 2023 09:43:04 No.7387262 Report

Quoted By:

>>7385672
>>7385760
retards I assume

Anonymous

Anonymous Tue 23 May 2023 16:57:37 No.7391449 Report

Quoted By:

>>7382558
>>7382562
its tiktok big brother

Anonymous

Anonymous Thu 25 May 2023 12:57:49 No.7394315 Report

Quoted By:

>>7382540
Farewell, /hdg/ #324.
You had a long illustrious run of six days and nearly 1200 posts.

Subject

Name

E-mail

Password

Capcode	All Only User Posts Only Verified Posts Only Moderator Posts Only Manager Posts Only Admin Posts Only Developer Posts Only Founder Posts
Show Posts	All Only With Images Only Without Images Only Spoiler Images Only Non-Spoiler Images
Deleted Posts	All Only Deleted Posts Only Non-Deleted Posts
Ghost Posts	All Only Ghost Posts Only Non-Ghost Posts
Post Type	All Only Sticky Threads Only Opening Posts Only Reply Posts
Results	All Grouped By Threads
Order	Latest Posts First Oldest Posts First

On these archives

On these boards

Your latest searches

/hdg/ - Hentai Diffusion General #324