Previous Thread: >>8871136>LOCAL UIForge Neo (ANIMA SUPPORT): https://github.com/Haoming02/sd-webui-forge-classic/tree/neoreForge: https://github.com/Panchovix/stable-diffusion-webui-reForge | Stable commit: 6964cebComfy: https://github.com/Comfy-Org/ComfyUI | https://comfyanonymous.github.io/ComfyUI_examplesSD.Next: https://github.com/vladmandic/sdnext>ANIMAhttps://huggingface.co/circlestone-labs/Animahttps://thetacursed.github.io/Anima-Style-Explorer>NOOBAI-XLhttps://civitai.com/models/833294/noobai-xl-nai-xlhttps://huggingface.co/Laxhar>NOOBAI SHITMIXES102d custom: https://civitai.com/models/1201815?modelVersionId=1491533291h: https://civitai.com/models/1301670/291h>PAIDhttps://novelai.net/imageNOTE: Do not feed the trolls. Don't take discussions about NAI too seriously.>RESOURCES4chanX Catbox/NAI prompt userscript: https://rentry.org/hdgcbEasy booru tag export userscript: https://github.com/Takenoko3333/Danbooru-Tags-Sort-ExporterTags: https://danbooru.donmai.us/wiki_pages/tag_groups | https://danbooru.donmai.us/related_tag | https://tagexplorer.github.io/#/Inpaint: https://files.catbox.moe/zy2xlr.jpg | https://huggingface.co/Wenaka/NoobAI_XL_Inpainting_ControlNet_FullUpscalers: https://openmodeldb.infoBooru: https://aibooru.onlineControlNet: https://rentry.org/dummycontrolnet (OLD) | https://civitai.com/models/136070 (look at the green links in the description)>TRAININGGuide: https://rentry.org/yahbgtr (WIP)Anon's scripts: https://mega.nz/folder/VxYFhAYb#FQZn8iz_SxWV3x1BBaJGbwsd-scripts trainers: https://github.com/67372a/LoRA_Easy_Training_Scripts (check out refresh branch) | https://github.com/derrian-distro/LoRA_Easy_Training_Scripts | https://github.com/bmaltais/kohya_ssOther trainers: https://github.com/Nerogar/OneTrainer | https://github.com/tdrussell/diffusion-pipeOP Template/Logo: https://rentry.org/hggop/edit | https://files.catbox.moe/om5a99.png
Request anchor!
>>8876295I request you jump off a bridge
highlighterinos?
Nyo way.
>>8876298here you go, finest jumping off bridge for you saar made in India™
childs cpregnancy!!!!
This thread isn't official until the highlights are posted.
>>8876311
thank god we get to see it again
>>8876317I had to repost it, bro worked so hard for it and it would just die in litterbox? Not on my watch!
>>8876311I like unofficial threads more anyway.
>>8876309in pakistan they would all be boys
>>8876295Requesting a nice pixiv name.
>>8876323xXx__pussySlayer69__xXx
>>8876314thanks for still posting gens, for what it's worth
I want to do a 4 frame animation. What's the best controlnet model to do specific poses with illustrious?
>>887632415 letters limit
>>8876323controlkek
>>8876327xXxPussyAFxXx
>>8876328Good one, but noodlekek can't make a pixiv, he'll be sued instantly. (for stealing)
>>8876327xMadSexHaverx
>>8876330what about pforpinnacle if you're a pinnacle poster
>>8876323ok i have another good one: sid the slop
>>8876323SuckemFuckem
poppie_merchant_nai
>>8876333I'm extremely confused, who the FUCK is pinnacle?
>>8876337wait
>>8876337its like a normal pinn but its got an acle or something probably
>>8876337what about uncle pinn
>>8876334It's going to be exclusively ART though.
>>8876342fartslip
>>8876344Ears too small, she's also kinda old. Disgrace of an elf all in all...
>>8876344could you make a shuriken out of elf ears?
>>8876342tesnickle if you have a patreon
>>8876342vonSlopdogh
its up>>8876357>>8876357
>>8876361You are not supposed to advertise it, you are gonna get refugees who made /hgg/ reach 3000 posts last thread (me).
Might make a hitler and anne frank LoRA then make them fuck to be honest
>>8876337dino \(dinoartforpinnacle\)
>>8876361>paki spammer is one of the main reasons we need another split>anon puts his slop as the opnibba what are you doing...
>>8876378Isn't she underage? =\
>>8876378isnt she like 100 by now? =/
>>8876378isn't there an unfunny joke about baking to be made here? =l
>>8876384obviously im not going to make lolishit, but she's 96 years old. shes in the ground but shes 96 years old.
>>8876388>obviously im not going to make lolishitgood goy
>>8876388Since she died 16, he couldn't have canonically fucked her any later thoooooooo
>>8876391ye but shes like a fictional character so just... put in the aged up tag??
>>8876391Were you there? Hitler might have been a necrophiliac you dont know.
>>8876393Non-canon is just not for me, you are on your own.>>8876394Okay but Hitler died a couple months later while she was still 16.
>>8876398Amazing pussy juice, wish I could separate it from the style though...Ear play is also very welcome.
>>8876398make him give the elf ear a blowjob
Clitty status=?
>>8876326Why is she so happy?
>>8876404because she just shrunk me down to ant size and is about to send me on a journey to the womb
What is she thinking about?
my clitty only trembles for bibici
>>8876406pork
>>8876406>noooo he just rimmed my friend nooo
>>8876407prooferino?
>>8876406I know what I'm thinking about. I'm thinking about the loras you are using that you haven't shared yet.
>>8876406>why is my torso so long>why does he have six fingers
>>8876412https://civitai.red/models/2472006/anima-rucarachi-styles?modelVersionId=2875414https://civitai.red/models/2444547/cassaruhshura-latest-style-or-style?modelVersionId=2848936
>>8876414Thanks, I hate both.
>>8876415Same, also can't believe they weren't already on dan/gelbooru
I just want Firefly I generate to look good, am I asking too much of anima?
>>8876420You might be, isn't that /aco/?
>>8876421No it's made by slant-eyes. I don't really care though, I don't play braindead videogames.
>>8876420whos firefly
>>8876423https://www.imdb.com/title/tt0303461/
>>8876423I was inspired by >>8873074But my fireflies looks like shit. Strangely enough all the other girls I gen look good in my styles, but not her, she fucking sucks??
>>8876427god tryna tell you somthn
>>8876413box pls anon very beautiful
The muscle mommy to normal mommy pipline
That's a man with a pussy
>>8876436I don't like the faces he makes, but it's nice otherwise
Anyone training Loras on AMD for anima? Any tips? Can't get any working at all.
>train the TE on an anima lora>it works perfectly fineさすがでね。
>>8876437stfu pedo
I'm new to using this "anima" thing but it seems to be pretty easy desuhttps://litter.catbox.moe/1sqsc50kyl8lfnum.png
>>8876446The question is whether it works better than the same lora trained without TE.
>>8876449>I'm new to using this "anima" thing but it seems to be pretty easy desuno shit. who has actually found anima hard to use? other than civitai retards of course
>>8876443i don't have any tips for training on amd but i wish you luck anon
>>8876447stfu trannylover
>>8876450kind of seems so, actually? doesn't seem worse at least.<https://files.catbox.moe/njxix6.png><https://files.catbox.moe/2izc2k.png>same seeds, same prompt, loras trained on the same dataset with the same settings/lr, only difference is the first has the TE enabled and the second does not.not that this is an amazingly reliable way to test anyway, since these lora settings aren't deterministic and I could run 10 of each of them at 1:1 and each would be slightly different from each other.also as an aside, whoever said that anima can just gen higher resolution as base t2i was actually correct. And this is kind of important because otherwise on flat colors you get a lot of dimpling type artifacts when you upscale(or at least, you do at .5-.6~ denoise; at .8+ it's reduced a lot but the composition also completely changes and it's pointless without controlnet tile to reinforce it)
>>8876460That's cool. Maybe it's something that changed since preview1.>otherwise on flat colors you get a lot of dimpling type artifacts when you upscaleBut do you really need to upscale flat color? There's no more detail to be added.
>>8876460training with te seems pointless honestly> And this is kind of important because otherwise on flat colors you get a lot of dimpling type artifacts when you upscale(or at least, you do at .5-.6~ denoise; at .8+ it's reduced a lot but the composition also completely changes and it's pointless without controlnet tile to reinforce it)more training at higher resolution will probably fix this with the later previews. this was a massive problem at just 1.5x on preview 2 and it is way less noticeable on preview 3
>>8876427Do you like this style?
>>8876465Not really, the pic I linked I liked the bloom and it reminded me of ke-ta sort of. Except @ke-ta is garbage on anima compared to noob.
>>8876467nta but he draws pussies nicely. i'll probably bake a lora for him later
stop genning firefly with ugly bastards ffs
>>8876460also forgot to add about genning higher resolution as base, one downside to keep in mind is that it tends to prefer making things skinnier when you do it for some reason. Which means you want to aim for a bit more of a fat aspect ratio.>>8876462>But do you really need to upscale flat color? There's no more detail to be added.Yes and fuck you for suggesting otherwise.but really it makes things look a bit more clean.>>8876464I mean, it's another nob to turn. Arguably it should be better for training custom tokens the model doesn't recognize. I just hadn't tested it and wanted to.>more training at higher resolution will probably fix this with the later previewsI actually don't think it will. I did a bit of testing and it's very directly linked to denoise levels on upscaling. At 0.6 denoise it's there, at 0.8 denoise same seed it's gone/reduced, and genning at high res at t2i it doesn't exist(though it WILL randomly fall back into weird diffusion noise patterns occasionally). Which suggests that the only way to "fix" it is is with controlnet tile and upscaling at high denoise. It's not really a "model can't handle the resolution" issue, it's a "model doesn't like retaining pixels and making a larger canvas" issue.
>>8876468Here you go
>>8876471>I actually don't think it will. I did a bit of testing and it's very directly linked to denoise levels on upscaling. At 0.6 denoise it's there, at 0.8 denoise same seed it's gone/reduced, and genning at high res at t2i it doesn't exist(though it WILL randomly fall back into weird diffusion noise patterns occasionally)you can clearly see an improvement with this problem between previews 2 and 3 though. you can test it yourself
>>8876472>>8876467Oops
https://files.catbox.moe/da4srl.png>>8876474I didn't really test v2 at all. All the testing is on v3 where it exists and it's a big fucking pain in the ass. And going back and checking what little I did with v1 it was about as bad as v3 is now.
>>8876476>miyabino_(miyabi1616)_loha_Anima.safetensors>578.81 MBWhat the fuck?
>>8876481And?
>>8876472Thanks, I tested this and decided on a radically different approach.
>>8876488gulp gulp
>>8876460>>8876471Have you tried tiled upscaling already? And just keep your tile size below 2 megapixel. You're right about what causes it, but it only happens if the input is too large and this is an easy workaround.>it's a "model doesn't like retaining pixels and making a larger canvas" issue.You make the larger canvas, then pass it to the model. It doesn't know that the input is an upscaled gen, just that it's big.
>>8876493These gopro loras get progressively uglier with time.......
>>8876480Awesome. When are you gonna share your newest lora with us, firedotbased?
more than half of the time you use "heart" with "fertilisation" or "impregnation" anima spergs out for some reason and tries to fit them both together like they're the same thing.
>>8876496Try placing the heart somewhere specific?
>>8876497i have no idea how to do that, im a turbo newfag
>>8876496artist?
>>8876503it was made with AI
>>8876503it was a blend: Mikoyan normal weight - low weight (.3)grimgrim
>>8876509thanks e
cucks
>>8876510Taк и нe выpeзaл шecтoй пaлeц, гoвнoдeл
>>8876493make her younger and pregnant
>>8876510feetbox? nice
This is the 2huai OP in May 2026, who is going to go and teach them about the new models?
>>8876510grippersbox
>>8876488sketchbox
>>8876521It's i2i with [@watanabe yasuaki|@asakuraf|@fkey|@fkey] (0.5 denoise euler cfg++ 1.0 cfg)but the starting gen is NAI: https://litter.catbox.moe/k0ax3f.png
>>8876522oh nyooooooooooooooooooooo~
>weather reports said that hot days were going to end yesterday>it was a lieaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
>>8876519why would you even want to? have you seen how bad that place is
>>8876518>>8876520https://files.catbox.moe/y3ruhr.png
>>8876529mankai kaika?
of course it's nai.nai is the best after all
>>8876530yeah
>>8876532he is a pretty based artist, I wanted to create a mix around him but it never looked good enough (for me)
>> 8310603
>>8876533havent tried mixing him myself, i just remembered he existsbut here's mankai kaika + hiramedousa, i think it looks nice
>>8876529>artist_collaborationNew anima snake oil?
>>8876542nah
>they just had to prompt it to mix artist all alongWhoa..
retard
>>8876436box please?
https://files.catbox.moe/8j92xg.png>>8876491I haven't done shit with tiled upscaling since sd1.5 days. And I've never been one to do stupid scale upscaling anyway. Either way your example doesn't inspire a whole lot of confidence since it still has a lot of visible noise.>>8876495lmao I'm not uploading an autistically tuned lora of my own style.maybe I'll do a 1dim at some point and upload that if it's mediocre enough. otherwise you can deal with bakaschizo's bake.
>>8876572>mao I'm not uploading an autistically tuned lora of my own stylepretty sure he was being sarcastic dumbass
>>8876572Doesn't have your artifacts though
>>8876572>otherwise you can deal with bakaschizo's bake.His loras are better anyway. Why call him bakaschizo though?
>>8876426I will never not be mad at Fox for how they handled it
>>8876609he's brown.
>>8876612Do you really think nobody will figure out that you're Indian just because you keep calling everyone brown?
>oekakischizo does stuff>seething followswhat causes this
>>8876614must be the smell
are we really pretending bakaschizo isn't brown now?
i'm making beef stew
>beefbro wtf man
>>8876618Why would you eat beef?
>>8876618you're not funny. how would you feel if i said i'm making jesus stew?
>>8876617I'm pretending to know who that is, so
>>8876620he's white
>>8876406He has melanoma.
>>8876635oppai wa dame
disgustingget those bags off her
>>8876635head too small. breasts too big
>>8876639this, give her huge thighs instead
>thighsgo back to l eddit
>>8876635pig- i mean peak
>great gen followed by a paki shitpost
guess we know who the obsessed anti-oekakibro schizo is
who
xis greatest enemy is ximself....
>>8876651I don't get it
>>8876652pakis fuck goats or something idk
>>8876653lmao
>>8876651As north caucasian I prefer sheep
>>8876651haraambox
https://files.catbox.moe/bwxytk.png
>>8876657>>8876660
those gens look as if it's a llm that was tasked to make basic ai-generated slop to fill a hentai thread
nah, llm would make better gens
oekakibro is now officially my favorite poster. noodlefriend needs to step it up.
I wish he could step a few shades lighter than brown..
shut the fuck up goatfucker lmao
Can anyone explain what's with the brown people self-hate itt?
common paki meltie
brown didn't like that one
@grok are pakistani brown?
Mmm.. pakistrami
you ever open up the UI only to realize there's nothing you want to gen
>>8876675Yes
>>8876675wildcards and generate forever all day everyday
brown behavior
calling everything brown still won't magically turn you white bud
>>8876572> lora of my own style.nigga you trained a low dim lora on unarsite and then started to draw over the results, its hardly your own style
he was balls deep inside a goat with tears flowing from his eyes as he typed that post
>>8876686try not to kill yourself ayakon lol
furious brown hands assaulting that poor keyboard..
>TI saga completePretty fuckin' neat>Remembers that CLIP doesn't exist in Anima, so I still cant use TI's for mixing like I used toAww... fuck. Does CLIP really screw things up worse than it letting us use scheduling? Are there newer or other options?
>>8876692make a patreon or whatever so we can fund your own experiments
>>8876692Are you asking about TIs or scheduling? Started mixing up terms there near the end.
>>8876694Not him>>8876698Sorry, can't use scheduling *ON* TI's in Anima
>>8876703Well, TI training seems broken on the easy-script fork, or doesn't support anima; and Civitai doesn't have a single example. But I would expect wcheduling it to work normally on Comfy, there's nothing architecture-specific about TIs.
>>8876686kek that style is like oekaki schizo's but without looking like ugly vector trace
>>8876707why do you fuck goats?
new schizo take >>>/e/3086838
>>8876709read just the first sentence and i agree with him, preview 3 actually feels worse, it's fried, less coherent
>>8876709>Remember that Adetailer and things like Controlenet do NOT work with Anima
>>8876709>I recommend inpainting problems areas with wai-shuffle-noob
>>8876686>>8876707Ouch not like this, dot inc bros ;_;
>I take commissions for custom LoRA models so it is in my best interests to know what I'm talking about. Well, if Konan says it, I'm sold. Downloading preview 2 again.
just because some brownie grifter says it, doesn't mean it's not true, chuddie
konan the barberryan?
Thank goodness. Bluvoll has saved anima 3.https://civitai.com/models/2530372/slivanimix-anima-preview?modelVersionId=2932616
>>8876718>providing a familiar AI-generated Aesthetic into most seedsThanks, prompting for AI-Generated was becoming too much of a chore
>non stop jav lora bitching>it finally gets baked>https://civitai.com/models/2610780/anima-realistic-asian-style?modelVersionId=2931511>not one peepI hate you all sometimes.
>>8876718hell yeah brother more ai slop
>>8876720well it's because it's clearly not javRETARDwe were asking for real jav screencap lora, not a lora of some random asian foid
>>8876718What? I thought the Anima arch was dramatically flawed and not very suitable for further training, and its licensing was also very questionable
>>8876720>no jav anywhere in the title or description lol>jeetvitailmao
>>8876720I did tell you nobody likes JAVbut yeah it's that shitty format of porn vids, not just photos of japanese people
>>8876722>>8876724I saw none of you helping when anon offered >>8875929
>>8876726anon should've did it, not asked for datasetthat's not /hgg/ culture
>>8876725>I did tell you nobody likes JAVowari da..
>>8876726The lazy fuck that keeps asking for jav lora should provide the dataset
>>8876726if i could do it all myself i wouldn't be asking >:(>>8876725>I did tell you nobody likes JAVstill just you chuddie
>>8876729nigga making the dataset is the hardest part
>>8876731and?
>>8876726Would've. Lost all my hi-def uncensored JAV in a lightning storm several years ago... RIP
>>8876735>hi-defdoubt
>>8876730>still just you chuddieCount me in as well, to hell with 3DPD.
>>8876738ntrman lora when?
>>8876738no... who killed my mother...
3dpd is trash, bros
>>8876736That weird 2K-ish format they used when it was new. Which, at the time, was pretty fuckin' Hi-def
>>8876741don't care about what paki thinks
only indians know what pakistan is
>>8876744I am from caucasus
>>8876746шo тe eбyт кoзлoв, шo эти
>>8876747Tишe, cвинья
>>8876748кeк
>says he's from pakistan >seethes about indians when people keep mentioning it afterwards really not the smartest turdie around huh
india is my country sirs
Goobers
>>8876707Yeah he used to share 2 loras trained on that guy a while back and all of a sudden he had his 'self trained' style.
>he's still seething about oekakibro 5 hours laterholy kek
I don't dislike any regular posters here, I hope they post more and keep activity up
>>8876762>thread retard is surprised that people arent on /hgg/ all day on a sundayclassic
who is goatfucker quoting
why is he called goatfucker again?
>>8876766see >>8876651 >>8876653also why he's so angry at fire dot today
>>8876763thisbut I'm also one of them, so
i bobble my head in agreement
>>8876522i've never seen someone img2img a gen just to make an anime girl ugly
>>8876774you must not spend much time in /asdg/
who?!
what the fuck is asdg
>>8876782it's where I post photoreal img2img of anime girls
/asdg2/ when
How do I convert a novelai prompt to work with anima?
>>8876790just... write
I threw up a little
why is lil bro posting /aco/ on /h/?
>>8876798lil bro is hgg culture, chuddie
that peni is not going into her pussi
>>8876790@ before every artist, is the main thingif you have regional stuff then rewrite it into NL like "girl on the left does XY"
>>8876801anatomy is hard for virgin turdies, please understand
>>8876709>I recommend inpainting problems areas withy wai-shuffle-nooball i needed to see to not take this retarded nigger seriously
her head is huge
could be huger
>>8876805Fuck off with your racist bullshit.
>>8876807yeah but then it's chibi not pedo
>>8876810the real problem is the tits, should be flat
>>8876808how is that racist
>>8876814Dude... the N-word?
>>8876815yeah that's racist
>>8876815it's a descriptor???
>>8876805box?
really tired of the horsegirl cringe shit
>>8876820take bobblehead out of negs and it should look like something similar
huh?
>>8876822god should have put brown in the negatives before creating you
problem?
careful, he might shit piss and cum all over the thread again
you sound excited about that
we do love shit here in mumbai :D
proofs?
is a big ole bobblin head really so bad?
only if ur a paki
i just bought more land in the novelaiversenow im gettin paid with aetherroomearth
bobblin head good boxing practice. pow! pow!
>>8876820hmm.... nyo
can you make her less extremely eye searing fried
>>8876843god shall handle you
>>8876844It's just a bright colorfried is something else
>>8876846you should learn to take criticism instead of making excuses. we are trying to help you
i like fried rice
>new bobblejeetgrim
it's always the colorblind ones with the ego
>>8876844prescient post meant for >>8876848
having a big head means big egoso it makes sense
bobblers wish it meant you have big cock
any lora bakers here actually use diffusion-pipe? Why is it so slow compared to sd-scripts for me
>>8876848Her CFG?
>>8876860over 18, sir.
>>8876860max of course
Babe babe wake up!, Laxhar Lab has reverse engineered and open-sourced the trainer for SenseNova-U1, a leaked unreleased image generation model.>Key highlights:no VAE or separate Vision Encoder, making it a true end-to-end text + image model8B parameters, making it fast and efficient compared to rivals like GLM-Image (16B)Strong at understanding complex prompts and generating infographicsTrainer is live on Hugging Face: Link: >>>/g/108791657
>>8876859are you comparing anima to anima?
>>8876863henti?
>>8876865nice, very anime
>>8876865Yes. Diffusion-pipe seems to also require more vram?
>>8876867what, cfg not high enough for you?
>>8876869can you share, the og gopro lora
>>8876872no :^)
>>8876873you are lonely because you don't give people much reason to love you
what do you need the gopro lora for? why do you hate anime?
>>8876878i love anime and that's why i need it to gen truthful Tifas
speaking of tifas, is there no ff7r lora for anima?
New lora for anime fans https://civitai.red/models/2613352/anime-style-loraanima-lora?modelVersionId=2934328
>>8876881real ff7 fans are using pony, tho
>>8876883why tho?
how green is too green?
>>8876884for genuine style, duh
>>8876888anima is way better at replicating styles, tho
>>8876889do you like trains?
>>8876881we already have the BLACKEDRAW lora
