Previous: >>109711423>UIsComfyUI: https://github.com/comfyanonymous/ComfyUISwarmUI: https://github.com/mcmonkeyprojects/SwarmUIKritaAI: https://github.com/Acly/krita-ai-diffusionForge Classic/Neo: https://github.com/Haoming02/sd-webui-forge-classic>Image ModelsIL: https://rentry.org/illustrious_checkpointsAnima: https://huggingface.co/circlestone-labs/AnimaKrea: https://huggingface.co/krea/Krea-2-TurboNovelAI (paid): https://novelai.net>Video ModelsH3: https://huggingface.co/Comfy-Org/MiniMax-H3Turbo lora: included above>How to generate anime:image : https://rentry.org/comfyui_guide_1girlvideo: https://rentry.org/wan22ldgguide (No H3 guide yet)>Style ExplorersAnima: https://anima.mooshieblob.com/ | https://animadex.net/Illustrious: https://tagexplorer.github.io>Checkpoints, LoRAs, Upscalers, & Workflowshttps://civitai.comhttps://civitaiarchive.comhttps://tensor.arthttps://openmodeldb.infohttps://www.seaart.aihttps://www.liblib.art/>Resources:Share Metadata: https://litterbox.catbox.moeImg2Prompt: https://huggingface.co/spaces/fancyfeast/joy-caption-beta-oneOnline metadata viewer SD/NovelAI: https://spell.novelai.devCatbox/Metadata Userscript: https://gist.github.com/catboxanon/ca46eb79ce55e3216aecab49d5c7a3fbOutput cleanup: https://rentry.org/RemovingDiffusionGunk , https://www.mediafire.com/file/vipr23exc5htmnt (batch processing python script)Inpainting Guide from an Anon: https://files.catbox.moe/fbzsxb.jpg >>106520607>/adt/ logo:https://files.catbox.moe/pakfgt.png>Beware inorganic activityhttps://rentry.org/animanon
>>109726175Based
>>109726164why is there western art in these threads now and why put it in the collage?
>>109726175Based. We need more NAI genners, where's the well of NAI genners at? /aids/? Gonna shill /adt/ over there.
>>109726164https://strawpoll.com/GPgVYE6wpnaWe know that this proxy ban evading schizo who is obsessed with Ani and also speed baking /adt/ botted this poll, but if any of you want to collaborate to remove the schizo drama offtopic rentry from /adt/, please vote YES. Do not let the schizo win.
(thick thighs:2)
>>10972637019 anons for /adt/? Nah, I don't buy it. I would expect at least 10 different avatarfags, or at least 5 if it were a prude general, but there are only 3 or 4, closer to the number of Yes votes.
>>109726370have you considered he is using a bot for op bakes? It's always right at post limit rather than waiting until it gets saged a few pages
>>109726370>>109726420>>109726501It's very suspicious how much effort and mental gymnastics you put into defending some 4chan avatarfag, "anon". Not to mention an avatarfag who got kicked out of the comfy org offices for his behavior, and has admitted to samefagging and fudding anima (and left mountains of evidence that he does much more than he admits). I don't think anyone wishes to see this general full of "his" self-promotion, samefagging and spamming.
anima is so cool, so sharp
>>109726528Does the sword happen to have a portal or something? Why is it detached tho?
>>109726522regardless of what kind of fixation you have it's not about what you schitz about, it's that you are schizing out in our general. go to /ldg/ if you are an avatar troon troll
>>109726522Both of them are doing it even in the other thread, it will never end they spent years doing this. Just mention Ani's obsession with sucking dick and watch him go nuclear because he's too stupid to not post shit using his avatar.Whoever updated the rentry needs a gold star.
>>109726522>>109726501if it helps to keep the schizo here, i don't mind
>>109726538This new tactic does more harm than good. But keep going queen
>>109726540>Whoever updated the rentry needs a gold star.What's the latest update? The rentry seems pretty chaotic at this point.
>>109726548It added his /d/ posting history which is very disturbing even by /d/ standards and the fact he connected his avatar to his actual identity.
why rapid fire seethe like that?
I've been out of the anime gen game for like 4 months now, what's the latest meta? Still anima? If anyone has a good 1girl high quality, high res wf to catbox, appreciated. Just pulled comfy so I imagine I'm going to need to spend a few hours sorting that..
>>109726556It's pretty disgusting, but I think that's probably the norm for /d/.
>>109726537her kimono is the portalisn't that nice
>>109726573krea2 is really good at official art for popular series, people are training anime and NSFW loras. It's a fun model and way smarter than anima
>>109726573Anima is no longer being developed, anon, I recommend you switching to other models like Krea 2 or Illustrious.
the relentless inorganic anima FUD is actually just exhausting at this point. it's borderline spamming. please just shut the fuck up
>>109726585I just find it hilarious he still post shit like that despite attaching both his name and brand to that avatar. Why is he surprised he can't get work or collaboration with any serious groups?
>>109726725This thread is inorganic, anon, but it serves a purpose: letting Ani scream into the void.
>>109726772why would ani be fudding himself tho? the only one screeching is catjack
>>109726772Not really he still has time with the other schizo to shit up /ldg/. They will do everything but post in /sdg/ despite one of them being a former paid shill for stability and comfy.
>>109726788Got any proof of those claims instead of trying to deflect from documented behavior?
>>109726813show us your id to verify
>>109726788>>109726794You need to act like a normal anon, post some slop anime pic, ask about Anima, and then wait for him to FUD, he express himself, he happy, he is contained here, he thinks he has achieved his purpose. It’s so easy when you think about it.
>>109726672>anima is no longer being developed, so you should switch to a 3-year old dead sdxl finetune (which is also no longer being developed)
>>109726832He really isn't he still seethes in /ldg/ both him and the /sdg/ schizo are in it for the long haul to the point they are going full swing in misinformation campaigns for months.They think if they lie long enough enough newfags would feel bad for them which doesn't work because all of the sloppy shit they have done over the years is so hilariously bad and retarded it makes even the slowest anon mark them as untouchable.
>only 7 imagesYou guys are pathetic
i dont get it. if jenkum and mugen and all that is so good why isnt he posting gens from them? seems like thatd be the best way to convince anons
>>109726370Why do you make a poll just to not accept the outcome? I don't like anis shilling for his product and samefagging so i voted no, why is that a problem for you anon?
>>109726851Illustrious based:>https://civitai.red/models/2904237/wai-simple-illustrious, >Aug 31, 2026>https://civitai.red/models/1359594/wai-mature-illustrious, >Sep 2, 2026https://civitai.red/models/1318945/one-obsession?modelVersionId=3218603 >Aug 11, 2026NoobAI based: https://civitai.red/models/1217645/sih?modelVersionId=3249278 >Aug 21, 2026
>>109726900develop deez nuts
>>109726900WAI and EIH returned? Neat
>>109726884He knows his image is toxicHe only wants attention for himselfOther than giant dicks he wants to suck, he can't gen anything to save his life
>>109726900Those are lora merge shitmixes.
>>109726900>i'm not burning the food, i'm developing it!
>>109726929>>109726175
>>109726945Your hands are fucked paypig anon
>>109726928And this ones? :)
>>109726960Don't you get bored with this?The model isFREE
>>109726929sluts
>>109726977Thanks for respect his signature resolution
>julien
>>109726968Same as my rent :^)
>>109727051I would assume as much because employed individuals don't care about such things.
I was poking around anon's pic >>109724723, what the fuck is up with Minato and water? Are the naked/swimsuit fanarts in the training that strong? Is it leak from the word "aqua"? But I don't think it's really that because most aqua tags refer to the color, not water.
>switches to ungoogled-chromium to escape from inpainting halo hell in webUI>it's still hereFUCK
>>109726573anima
>>109726960The aesthetic versions are full finetunes.Turbo is probably multiple loras merged, but they're custom trained and not just shitmixing other people's models.All of the recent aesthetic and turbo versions involve reinforcement learning (what other local baker even knows how to do that?).
>>109727133Lol schizo
stocking at the train stationwhere will she go?
>>109727284>what other local baker even knows how to do thatRelax, he'll remain the sole guy who can do it, and he's taking that knowledge to the grave, since he won't even share how to finetune Cosmos itself with anyone.
>>109727302BLACKED
>>109727310>he won't even share how to finetune Cosmos itself with anyone.He has shared a trained lora, the dataset used for it, and the training config file. If you're full finetuning instead of lora, you only need to adjust learning rate down, that's it.But you already know this.
>>109726370Did you expect users of the contrairian website to agree with you? Maybe next time don't make your bias so obvious. >>109727354>But you already know this.He doesn't. He's never finetuned before.
>>109727630He's also never done anything in the animation space but will name himself after it....keeeeeeeek
>>109727354>>109727656Neither of us wants to keep going in circles on this but the forgetting problem is real, it's documented, and it's why people are moving to Krea, or Anima with more layers (which exists specifically as a response to that flaw) or Mage Flow. And as of today, September 4 2026, people can't bake new concepts into Anima base. Whether CircleStone intentionally shared his loras training settings so loras built around his base model work but he didn't shared how to train his base model that's a separate conversation.
>>109727736qrd: ani was right about everything and catjack lost her mind because of it
So what dataset and architecture does Ani want to train on anyway?
>>109727736>Neither of us wants to keep going in circles on thisNo, I will keep going in circles on this because none of these arguments make any fucking sense at all.>the forgetting problem is real, it's documentedWhat's real is that people are retarded and are frying anima with bad settings and poor training practices. The model doesn't suffer from forgetting (at least not more than any other model), nor does it have any fatal architecture flaw that prevents training. How do I know this? BECAUSE IT WAS TRAINED. Anima itself is a finetune from Cosmos2! How could it be trained, if it's untrainable? You think the moment the last base model training step happened, the entire model just magically stopped being trainable? Like this is genuinely the dumbest argument I've ever heard. I'm supposed to believe that the model finetuned from Cosmos2 is actually completely unable to be finetuned.>why people are moving to KreaThere is exactly 1 Krea finetune: Kroma, and it's forgetting the base model's knowledge the more it's trained, more than any attempted Anima finetune ever did.>Anima with more layerspure snake oil, happening because people are unfortunately believing the "untrainable, it just forgets" FUD>Mage Flow1 guy in the noobcord has spent $150 doing a test finetune and it looks like complete ass>he didn't shared how to train his base modelYou take the config file he shared and reduce the learning rate, sweeping a few different values to find out what works best for your particular dataset. BTW, Krea, Ideogram, the Flux models, Lumina, SD3, NONE of these share any details about training whatsoever. You are trying to hold Anima to a higher standard than any other model, and despite this the dev actually has still shared training datasets and configs.
>67/12We need NovelAI genners
>>109727847Big "female" dick he can "service"
>>109727889A couple points though, since you clearly put in the effort.The elephant in the room is that the dev has never actually shared the true base model training settings, just derivative lora finetune configs. And training base models today isn't the same ballgame it was in the SDXL era, the techniques have moved on, so reverse engineering it from a shared lora config and a learning rate sweep isn't the same as knowing what's going on under the hood.And at the end of the day, the fact remains that people are moving away from that model, en masse. You can chalk that up to bad training practices and skill issues if you want, but that means you're arguing an entire community, Civitai, Discord, the finetuners, are all incompetent at the same time, on the same model, in the same way.Anyway, I'll leave it there.
>>109727948Kek based
>>109727979
>>109728082He posted that recently it's in the OP under his /d/ post. Still trying to figure out the decision making when you attached your government to that character and software.
>>109728114okay
>>109728090no, the kind of tuning the base model got is akin to foundation model training not fine-tuning since he had to overwrite most of the data in the cosmos model. It's a different process
>>109728156>he had to overwrite most of the data in the cosmos model.So it forgot?
>>109728156>training a model on millions of images isn't a finetune>a lora isn't a finetune>a 10k image finetune isn't a real finetune, that's a baby datasetwhat the fuck is a finetune then? ideally, give some specific examples
the real blackpill is that there is no such thing as a 'finetunable model'. illustrious forgot everything base XL knew, chroma is a melted bastardization of flux, etctraining cannot introduce new concepts without destroying old ones, the same reason loras are so shitty and force the character/concept's attributes all over the scene. if you want a balanced finetune, you would need the entire original model's dataset as well
>>109728090Kek
>>109728240>illustrious forgot everything base XL knewYes, but Anima absolutely did not forget everything base Cosmos2 knew. It is surprisingly competent at general-purpose realism prompts, it's really just the textures that are off. Which is why the whole "catastrophic forgetting" meme about Anima is extra retarded: it has retained the original base model's knowledge BY FAR the best of any anime model.
As for the Anima defenders, everyone's going to eat the soup they made for themselves. Each person will realize it in their own time, at their own pace, that Anima base 1.0 is already obsolete. But that's a personal maturity thing, everyone gets there when they get there.
>>109728383I don't really care about anima and think Krea2 is a better model, I just don't appreciate some literal faggot shit talking and spreading misinformation because the guy he wanted to fuck him in the ass wasn't gay and decided to fund someone else that wanted to provide a model for lesser hardware.I will insult him for using a shit tier TE that has crippled the model's room for growth in the future
>>109728383This. Just subscribe to NovelAI already
>>109728383Funny in a sad way how it seems like every finetuner nominally abandoned the Anime base model and switched to using the extended layered versions instead or Illustrious
>88/18Gens? Where is the Astolfo straggot?
>>109728458nobody is 'finetuning' anything you retard, local is stagnant
>>109728465you talking about me?
>>109728383Man why are you such a spastic homoIt's a model anonFor generating imagesYou can switch models however you likeWhy does it seem like you have some personal vendetta against anima and/or tdrussel? It's not normal to seethe for months about something free which you can just decide to not use if you wish
>>109728465i have five more of bocchi falling if you want them
So we're just doing this now?Well, if in France...
>>109728686I'll take two but can you maybe solo, panorama, high angle,facing_viewer, standing, winter coat, in snow, outside, deep snow, full_body, panorama, bocchi_the_rock!, playing_guitar,
>>109728571Yes, thanks for pksting gens ^^
>>109728702
for some reason bocchi has no eyes unless looking_at_viewer
>>109728595Better that this insufferable faggot stays contained itt instead of shitting up other generals
>>109728810But I like Anima tho, why do I have to deal with some schizo fudding it 24/7?
>>109728838There are worse things in life, kid. I have to deal with neighbors who blast music all day long.
>>109728794>>109728799ty the top looks like the ai wasn't sure to make her look up or forwardIt might just be the prompt I posted
>>109728892this is with looking_ahead
>>109728810He's not contained here, he's shitting up other threads
>>109729022NAIce
>>109728972She's so smolThe details on the falling one came out better I think, ty for trying but the falling one was nicer desu
>>109729097:o>>109729162don't mind if I do
>>109729093if you're going for that here's the gen with the same artist tag as the falling one
>>109725699>Anime diffusionis doomedThe fuck are you talking about? We're literally at an era defining moment where gens are near indistinguishable from seasoned artists. You only don't realize how good we have it because you're in it. Looking back this will have been a moment of history.
>>109729022lol I think genshin fashion confuses model enough that completely_nude (minimal prompt, Full model) needs to be pushed to 1.5-2.5:: to work for this character without adding "nsfw"
>>109729198I personally like this one a lot more
>>109729223For real, NovelAI has pushed anime diffusion to an absurd level.
>>109729192
>>109729502can you gen them getting divorced
>>109729502>>109729192>>109729097
https://huggingface.co/RicemanT/MageFlow-Stuffs/tree/main/epoch30Neat, an anime finetune based on Microsoft’s MageFlow with a 40K dataset. Looks like a good concept model.
>>109729622So, in your fudding schizo mind, this is a good model but Anima isn’t?
>>109729648lol, I’m not *that* schizo, it’s a good model, desu. With only 40k images it learned the style surprisingly well. Remember that anon who wanted to make a finetune from Z Image?
>>109729502>Kita you smell sweat and the boys toilet>I don't like your hair>Also, Nijika has bigger boobs
https://litter.catbox.moe/nbl8x17b4yey1gm8.pnghttps://litter.catbox.moe/by8s3k64jz9ny7y1.pnghttps://litter.catbox.moe/eeskjhhqcfki05mi.png
>>109729844ok, the dark charcoal? is a vibe>>109729909bless you
too risque for evening attire?
>>109729941R-rape!
>>109729951'It's' not hot outside, she's the hot thing
>>109728090bluvoll, selpah and Anzhc literature confirms that the model cannot be trainedthis does however ignore the fact that all three of the above are totally clueless and larping
>>109730091>>109729810Very cute :3
>>109730168/adt/ should build its own ideology and politics, separate from /ldg/. Otherwise what's the point of /adt/ if it's just an extension of it? Ani being a contrarian to /ldg/ and leading here makes things way more interesting. I don't want /adt/ to just be Catjak coded.
>>109730204Is this some kind of noble phantasm from fgo?
>>109730186ty :) https://files.catbox.moe/rdgtj0.png official style,anime coloring, (anime screencap) seems to be it
>>109730228>adt/ should build its own ideology and politicsbut why noobcord?
>>109730204for ai, I think it's very very good, local or notIf you'd tell me this was from a ( indy) gacha game I'd prob believe you
>>109730248Noobcord's vibe is kind of the opposite of /ldg/'s in a way. They have taken in Ani as a refugee, and they are not that into Anima, tdrussell, or Comfy either. That actually makes for a decent starting point for /adt/ to build its own identity as a thread, even if it's a bit of a borrowed one.
>>109730204Good, this gives me some kind of Wuthering Lane or Azure Waves vibes
>>109730278I think it's funny Lax works for comfyorg and most members hate comfyui
I notice V5 defaults broom riding to sidesaddle or some variant of not-between-the-legs, which avoids the "pole going through a dress" error. While V5 is an "improvement" in this respect, it still assumes they don't just sit on their dress.
>>109730286And bluvoll hates Lax, they are troons.
>>109730186It is actually this lora I am a dumb https://civitai.com/models/1826946/nanase774938-style?modelVersionId=2067471
>>109726164i have a ryzen 7700, intel b570, and 32gb ddr5 ram what models should i run to generate decent anime images and goon to feet rp is that even possible i know fuck all about this stuff
>>109730291V4.5 is a genuinely great model. V5 breaks anatomy, feels unstable, and the good results I've seen are mostly lucky seeds or tag vibing off an overused, overfitted dataset the model can easily fake its way through. Complex scenes with Madoka or Bocchi are a good example.>>109730321Ryzen is AMD?
>>109730228>Otherwise what's the point of /adt/ if it's just an extension of it?t. doesnt remember that the first handful of adt threads were filled with reposts from ldg in an effort to kill ldg kek
>>109730326yeah it's AMD
>>109730328>were filled with reposts from ldg in an effort to kill ldgBased
>>109730331Mmm, I believe there are WebUIs available for AMD, like ComfyUI, but from what I've heard, it's fairly complicated to set up. Not sure why, but AMD and AI tools just don't seem to play nicely together.
>>109727112>>109727124That's because Aqua is maid of the sea!
Catjak broke him......
>>109730346just to be clear it's my processor i'm using an Intel GPU thoughever. if it's true even the CPU is a problem i guess i'll move on to other hobbies
>>109730328Schizo reposter here, kek, good memory. Look, /ldg/ is trash, sure, but my reposts came from Reddit, not /ldg/. So by the sacred laws of thievery, a thief who robs another thief earns a full pardon. That's just how it works.
>>109730328I can't say I'm surprised. It was interesting to see that when anima dropped, before any of the drama surrounding it, the only diffusion thread talking about it was /LDG/... it was weird to see the "anime thread" not paying attention to anime models...
>>109730358>3 copies of quality tagsholy copypasta
>>1097303674gb vram you can run anima and sdxl
>but my reposts came from Reddit, not /ldg/.but why would you lie about something anyone can look up in the archive?
>>1097303671 you can literally ask ai to give you a specific route to start, not to be a dick but it's really handy, it helped me with comfy2 I believe AMD used to be an issue bc they don't have cuda cores, but that has been fixed for some time already,a gain, ask Claude or smt, it knows, and I'm sure there are videos.If you have a GPU with 8GB VRAM you'll be able to get somewhere
>>10973035810::painterly::
>>109730391He's using an intel gpu which can work with stable diffusion models.>>109730367ask an AI for a guide on what model to use specifically and how to set things up.
>>109730391word i just feel brain fried today and i'm uncertain how accurate it would've been since this stuff seems to move fast thanks for the info
why does this make him so mad
>>109730382>>109730394How did you know all that I thought 4chan stripped metadata?
>>109730425nai pics still work
>>109730425Thanks for the cute prompts, NERD
>>109730405well you can make it as complicated as you want, there's no law in asking Claude "hey I want to use comfyui to generate fat foxgirl but I don't know how" and it will answerIt's completely doable for someone like me so someone like you will manage too
nooooo they stole my prompts...
https://files.catbox.moe/vyq59u.pnghttps://files.catbox.moe/fz7p8u.pnghttps://files.catbox.moe/lesrka.pnghttps://files.catbox.moe/midlqn.png
>>109730697She seems to be enjoying herself so I guess it's okay
>>109730358>-2::doll::cursed artist combo, also have to add poster_(object) to that negation because it pops up everywhere, makes me wonder wtf is going on>>109730635I first read that as stole my panties
>>109730635Ans your feet!>>109730501Cute :3
>>109730291V4.5 looks much better than V5. Look at the hands, the face, everything. It's horrendous.
>>109730734>Ans your feet!what does this mean
>>109727354>the dataset used for it,I would like to see the dataset - assuming it isn't Danbooru 1024. Does anyone know where it can be downloaded?
>>109730747I don't trust hands in either model for full body shots, and in this one V4.5 is slightly zoomed in and 28 steps vs 23 though I don't think steps usually matter much at 23+.
Someone wanna donate me 3k buzz for the Janku Krea2 anime checkpoint?
So does anyone know if it's best to fill in as many negative tags as possible or does not it matter?
>>109730356based
>>109730635It's all fun and games until they start spraying to mark you and your bed as their territory. You can't get rid of catgirl pee smell, it's hopeless.
>>109731073Gah! Wong image!
>>109731045Doesn't matter
>>109731020Janku the slopper made a Krea checkpoint? Holly based
>>109731077you meant to post that on /e/?cute pic but hopefully they don't give you a vacation
>>109730814Putting V5 through its paces right now, and I can confirm that even at 1600x640 full body shots, hands, toes, and nails come out looking like they went through a wood chipper, and that's with inpainting. Could be my character, could be my style, who knows but 4.5 could do full body shots at 1216x832 and just got it, nailing placement every time I inpainted. V5 swings for the details and misses spectacularly and the lineart looks like it was made during an earthquake.So either V5 is undertrained, or its sweet spot isn't 1 megapixel at all, but something much bigger, like Krea's 2K or even 3K range, and the model just can't compress that much information down without losing the plot.
>>109731020Lol it's true.
>>109731020Anima seems to be losing checkpoint makers one by one. WAI, One Obsession, Janku, Hassaku, Neta Yume, EIH, and Noobcord.Does anyone know why?
trani failed to secure funding 5+ times and never baked apache anima. Does anyone know why?
>>109731667why won't they let me just buy buzz with a credit card? why this crypto bullshit?
>>109731708Probably because Visa and MC denied processing payments for them
>This is a confirmed regression introduced in v1.47.3 by PR #12318, which refactored the MaskEditor save pipeline to composite layers client-side. The root cause is that canvas.toBlob() premultiplies the alpha channel, which destroys the original RGB pixel values under fully transparent (masked) areas — they get zeroed out to black (or produce the corrupted/solarized artifacts you're seeing depending on GPU acceleration) [1].>This has been fixed in PR #13758, which was merged on July 18. The fix bypasses canvas.toBlob() entirely and instead uses a custom PNG encoder (encodeRgbaAsPng) that writes straight (non-premultiplied) RGBA bytes directly into a PNG stream, preserving the original RGB data even when alpha is 0.I wonder if this is possible for WebUI?
What anima checkpoint is actually good with IL loras?
Suggest me some Anima shitmix to use>>109731820>using Illustrious LoRA with Anima-based modelsAre you retarded?
>>109732626yes?
>>109731820anima needs anima loras bwo...
The pride of bearing a red band across the soul
costume swap>>109732739damn right
>>109732739cum in ass please.
>>109732970this isn't /b/
>>109732674Cute hoser!
>>109732989'stolfo tummy&bulge uwu
In light of recent events I wish you all a very bock saturdayWake-up music https://youtube.com/playlist?list=PLALm9OqeAm6nAXAnBI-P1rJKfeVPuD23v&si=2A9QWJG3HuAJX1AF
Or if that’s too gay for youhttps://youtu.be/wwfGhbmpq1o
>>109733007That playlist rules, anon, boppin' to it right now. Bow-wow-wow. Relax, don't do it~ (when you wanna c*m)
>>109733152Core memory for me, and I’m not even bock-oriënted
>>109733211What can ya do? It's a great track. Here, let me share the song of my people: https://www.youtube.com/watch?v=PcRyjkYdDxM
>>109733263Sick charcoal painting. Love the sky's texture in particular. Also, is that ... a black hole?
>>109733267Has sovl thoughCozy
>>109733007Oh neat a cool playlist while genning anime girls(boys) :3
>>109733378>anime girls(boys)Based. You better share (or share where you share).
>>109733263V4.5 or v5?
>>109733273just the sunrise>>109733385v5, it should have metadata
>>109733397I mean your preference, which do you prefer?
Anyone know how to get ris running in Comfy or Forge? https://github.com/shitagaki-lab/see-through Looks like it could be a good way to get rid of inpaint halos.
https://files.catbox.moe/oihww9.png
>>109733706Hawt alcoholic minor Would feed booze till she starts drooling
>>109733706Cuteeee
>>109733805Lol
>>109731020Just wait, cui_ .First versions of those kinds of big projects are always trash. Also, it's a homemade fine tune on a very big model, there's a good chance the gains are minimal.
>>109733978>minorshe's a hag
>>109731667Raehoshi continues to develop Anima base and has shared xer training settings.
>>109731222Yes, sir, we are sorry for the unfortunate catgirl incident. Corrective catgirl disciplinary measures are being taken to ensure it does not happen again.
I can't get a spanking out of krea2 no matter what, so frustrating. If I get a handprint it drips red, makes her panties a soaked mess, or it ignores it. I'm sure the model knows how to do it too, but is prevented by safeteycucking, and if you block the safetycucking, it turns too stupid to follow the prompt. Aaaaggh!
>>109734706this is what happens without a booru finetune. krea 2 will forever remain a potential model
>>109734706anima works for me :^)
>>109734646 >>109734706 Krea 2 is doomed. I’m not sure whether your goal was to jailbreak the model as a challenge or to use it for that purpose. If so, the most practical and natural option would be to use Anima, since it was essentially made for that.
>>109734756Trust in Janku
>>109734330Hey! Don't call bullshit to my #84858 coomslop character of the month! Personality trait: cute feet, cool hairpins! >:^(
>>109734082IMO, the first versions usually do the heavy lifting, but they’re also more unstable. Subsequent updates tend to focus on stabilizing the model and making smaller improvements over the major changes introduced in version 0.1.
>>109734893I don’t get why anime became the default for gacha games. Are there any gacha games with /aco/style that also follow the formula of putting them in cute outfits and monetizing? LoL, Dota 2, or Path of Exile don’t really seem designed around that kind of appeal.
>>109734893I think the model messed up here and generated a blue archive-like character but not one that actually exists
@grok make anime girl big boob
