Previous: >>109439582NEW: https://huggingface.co/circlestone-labs/AnimaAnima1.0 is out>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>How to generate anime:image : https://rentry.org/comfyui_guide_1girlvideo: https://rentry.org/wan22ldgguide>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
I forgot if I already posted this one here
>1girlslop
>>109471633>NEW: https://huggingface.co/circlestone-labs/Anima>Anima1.0 is outisn't this like three month old already?
kill the old thread first retards. we already have the same schizo double baking in /ldg/, we don't need this shit especially when this thread moves so slow
>>109471798rebake and get rid of the schizo stuff too. you just invite them to stay
>>109471801we reached bump limit in the other?
>>109471810you let it get to page 10 at least. fucking hell
>>109471633I think we all came to the concencus you are fired from baking another /adt/ again! Rakun only position!!!
>same schizo double baking in /ldg/what is /ldg/? when i search /adt/ only this thread appears
just came back from the breakupdate comfyUI and it tell me to get cu130 something, I'm on 128Should I update? for 3090. Last time I update it break sageattention and shit
>>109471910I would not switch to cu130 just because ComfyUI is suggesting it anonnie
>>109471633Schizo bakehttps://rentry.org/ranfaggot
>>109471801>>109471808>>109471865>>109471881>>109471954you're not being stealthy
>>109471966like the metadata?
>>109471781The 666th March? That's a lot of Marches I must say.
thanks for bakering
debo love
>>109472213>no gen/ldg/ slut! >:^(
>>109471783how about 2girls
For me it's UNET...
>Comfy UI APP modewhy it flopped?
>>109472254i dunno if i can post the vids im making here
>>109471633>Important Anime News & Model UpdatesNovelAI (NAI v5 Rumors Multiple users mention that NAI v5 is rumored to dropTIPOv2 Release >109466756Illustrious Z 2.0 A web demo for Illustrious Z 2.0>Technical FixesForge Neo Metadata Issue >109451482 - >109452298Speed Boost for Minimax Video Gen >109465841 anon shared a method to cut Minimax generation time from 500 seconds down to 176 seconds using Sage Attention 2 + Spectrum.ADetailer SOTA Model >109462670 For face detailing in Forge Neo, the current model is yolo26l-widerfaceComfyUI Tag Autocomplete >109470031 To get Danbooru tag autocomplete in ComfyUI, users recommend ComfyUI-Autocomplete-Plus by newtextdoc1111 Saving Defaults in Forge Neo >109450886 To stop Forge Neo from resetting settings on reload, go to Settings > DefaultsStopping Clothing Bleed in Anima >109447728Preventing Random Background People in Illustrious >109459827 & >109459918Video Generation Prompting for anime >109466614
hand-holding is complicated...
>>109472452how about you inpaint in Krea?
>>109472482how about I paint the insides of your mum?
>>109472345Can you do the same for /ldg/ threads so I don't have to burn mental energy reading 80 posts of pure slop just to find something even remotely notable?
>>109472720Thats literally my entire months expenses counting rent spent just on anlas>>109472345appreciated
>>109472313they still can't even get subgraphs to work right
>>109472313Forced layout, can't select input or outputs inside subgraphs, no collapsible input groups, even the feedback report for the mode has typos, etc. Currently there is no reason to use it over subgraphs locked in place.
If you're trying to make a lora, but you have a small data set, so you duplicate and horizontally flip every image, will that help produce a better lora, or will it just fuck everything up?
>>109472313Forget that python bloatware over engineered piece of shit.
>>109473349If your character has asymmetrical clothing, I mean anything from a coat that buttons up on one side to an eye patch DO NOT USE FLIPPING OR MIRRORING FOR YOUR AUGMENTATIONS. You're feeding it conflicting information and the model will get confused and start generating mirrored artifacts!!
>>109471633lots of my preggo pics yaay
>>109473349Better zoom in and crop also you can flip non asymetrical crops
>>109471633Why not AnimAnon's gens in the OP ??? >>109439754
>>109473349Dumb question, but isn't the point for the lora to learn that grain texture? That kind of detail only gets learned properly at 1536x, no?
>>109473349You can also fry and de-bias it a la copkihttps://civitai.red/articles/33356/copita-train-an-anima-sdxl-style-lora-from-a-single-imagehttps://note.com/emanon_14/n/nb837cd8bec9bBasically make comparable datasets to what you do have ensuring the only difference is the aspects you wish to be trained.
>https://github.com/newtextdoc1111/ComfyUI-Autocomplete-Plusthanks for this anon btw, it does look like a lot of helpsomething else that is this useful and out of my radar?? I'm proompting with anima
>tfw have to wait for offloading to get fixed so I am just playing around with the ui display insteadI raped my gpu for this masterpiece lolhttps://files.catbox.moe/hsielx.mp4at least I broke the ui curse of being the opposite of what it's named>>109473556is this something you would like to see standard? the new text editor extension I am using has a feature for autocomplete like this
>>109473795cute!
>>109473827thanks. i actually forgot to switch the first and last frame but it worked out. she was supposed to be grinning from the start. guess i can make another gen to see how it looks with the frames fixed
>>109473834>but it worked outi liked it, made her look devious.
>>109473827Pedo
>>109471637cute af
I'm new to prooompting, how do you achieve clean, stylized text bubbles like this in ComfyUI Anima? I remember the person that made this picture they told me they were using NovelAI fon generations but I didnt ask anything else back then
>>109473795>no sudden jumpscare from the box
Error: No text entered.
>>109474131now i want to have kogasa pop out the hole
https://github.com/Haoming02/sd-webui-forge-classic/issues/1380
>>109474173>maybeeven ani has it. is this the end of auto forks?
>>109474127just add the balloon with the text after the image was generated
>>109474291unironically asking any nice tools for that? that have cool fonts and stuff
>>109474299clip studio paint (paid) or krita (free)
>>109474299If you really need text in the Gen, you can just use krea2
>>109473349Hm. It...kinda worked. Kinda. Sorta.
Hey guys, i created a blind artist tag ranker for NovelAi specifically, i did it for myself but i thought it would be nice to share.https://github.com/SeaBassBand/NovelAI-Artist-RankerI am aware that one already exists, but mine is more complete and for people that have nothing to do, i also updated the list of artist tags the previous one used since it missed a lot of famous artists.
so if I want to save prompts and stuff most people just drag back the image into comfy right, instead of having 10 notes laying around like im doing
>>109474798Most AI packages save metadata in some shape or form into the image itself, ComfyUI saves the whole node workflow so you can just drag the image into Comfy and have the same workflow (Unless that person erased it on purpose)You can also check an image metadata with tools like this: https://sprites.neocities.org/metadata/viewer
>>109474798Feel free to drag and drop this into your Comfy. Anima is good, but the other one is my own and it's garbage,Anima: https://files.catbox.moe/r71vgq.pngMy workflow: https://files.catbox.moe/t2h3zc.png
How did they manage to make the cow's udders this good? I tried everything to replicate this picture in Anima, but with this part of the animal it just makes whatever. Was it just another model? Any clues??
>>109471633Better to have 16 9 aspect ratio which will better use monitor space for most people while the phones have to zoom in anyway so it wont matter to themAlso after merging all images and saving it as jpg with below equivalentresult_img.save(out_path, "JPEG", quality=95, subsampling=0, optimize=True)You can compress it much more without much quality loss with jpegoptim --all-progressive --strip-all --max=95 1.jpg
>>109474898krea2
>>109475073shit I just noticed the duped hand
>>109474364Oh thanks! ^^ +1 kurumuz to you fellow NAIer
https://files.catbox.moe/3reh95.png
>>109473429When compiling a data set for a lora, do I want to copy to remove all white/blank/empty space, or would that just train the lower to produce images with no space between the character and the borders of the image?When compiling a data set for a lora, so I want to make sure all images are a power of 2, such as 1024x1024, 1536x2048, etc?
>>109474185>even ani has it>last commit one week ago>last release 6 months agoleast obvious samefag
RDBT is a starved attention whore, daily updates of the same schizo garbage. Blocked his retarded ass, go die somewhere else, you clout chasing faggot.
what are some anima artists to make this kind of 1girl cooler with better boobs?
>>109472345REAL NIGGA thank you
>>109476384>cooler with better boobswdym anon that's extremely vague and subjective lol
>>109476470yeah and I want some answers depending on who you arewho I am isn't working
>>109476384Try wanke, pigeon666, and p-nekor, those are my goats for NAI and I think Anima has them too, exact name may differ tho.I also love Sciamano240 but he has a more gooner art style, and well thats what I use his tag for. I discovered those tags using my ranker!!1!1!1!... Just saying
>>109476384what kind of tits do you like? there's plenty of artists that do good tits, but it's all up to your preference. i like derauea, akagi rio, yukimaru (gojo), raisuta, loliconder, tamagoroo, for big booba anime girls.
>>109471633>1/30 positive 4 protruding nips n sfw vageenthe pleb ratio is too damn high
>>109476680>>109476384Oh yeah I forgot, try also zis, azomo and tenobe, tenobe has good boobs.>>109475165Thanks :)
>>109474898well what exactly did you try and how were your results?
Some hit song
>>109476881it just wont do the udders, i tried increasing priority and adding anatomy tags and that, nothing works
i like the idea of big, muscular, heavy armor clad, buff, female characters, but i find the leaner aesthetic is really not too bad either.
>>109471633>R: 88 / I: 28Gens?
>>109476994
>>109476994i can't decide which one looks better
>>109476994help me pick one bros
NAIgods, does anyone know a better way to use their Anlas? I haven't even spent 1000 yet, and now it's getting harder to burn through them because with the Forge extension for NAI, where I can use img2img and adetailer with the NAI model, it costs me nothing to generate txt2img, upscale with Nvidia at x1.5, and then inpaint at 1024x...
>>109477337find a high quality prompt and burn it on the highest resolution? not much else to do
>>109477337speaking of NAI when will those fucking jews release 5.0. any news/rumors?
>>109475349https://files.catbox.moe/f3xv6x.pngRevised and img2img'd at very low denoise (LCM/Normal)
>>109476977anon, it's krea2 like Ani said >>109475073
>>109477148>>109477160>>109477167i think i would pick >>109477167 but they all look good
>>109477591thanks anon, hard to decide, sometimes its better to let others pick for you. i'll need to inpaint the face on that one, came out a bit scuffed
How do you train an AI to learn specific colors from the hex value? I'm having trouble getting the AI to break away from colors that stay within the homogeneous palette of the generated image.
>>109478055probably just by being extremely selective about the values. doing every conceivable hex is just way too big of a dataset. I'd imagine a select lora tags/captions for solid color kind of defining the hex code to a color then having examples with the same values and captioned for that. again, you's just have to get the values for whatever production or project you are doing and have separate loras/project
>>109476659Seems to be working to me, looks great. Are you using a particular lora?
Can anime models do zettai ryouiki cowboy shot properly, or do they still shit up the legs and make them slenderman like long?
Why is there still no proper controlnet for Anima?
>>109478523I haven't tried it but isn't there a Depth ControlNet available?If you need Canny/Lineart you won't lose too much by just using Illustrious since its weakness relative to Anima is detail and composition and Canny/Lineart will make up for that in a lot of cases.
>>109477337the way I used anlas in the past is id just let the sub run out then use my anlas before resubbing.
is julien here? i have a question
>>109478435Are you coming from sd 1.5 days or something? Proportions might depend on artist and resolution to an extent, but yeah cowboy shot zettai ryouiki is pretty basic.
>>109478543Anima's detail is too good. I can't go back.
>>109478834Just use depth cnet if it works
>>109477588>>109475073Why do you guys think it's krea2I barely see it used for animeCould in the future Anima v2 be trained with real pictures or something so it works better?
>>109478523>controlnetWhat does that do im new
>>109478435I imagine the answer is favourable as I don't even know what you're talking about
>>109477167>>109477148>>109477608it's immensely important for female knights to have their buttocks exposed to guarantee enhanced mobility during fights
>>109478847way smarter model and it knows official anime very well. I recommend it if you want to make fake screenshots of anime. it also does the slop ai look easily
>>109476977This is just what happens when you don't put autistically tagged furry content in the dataset but do put in deviantart.
yamero Stockingu you already ate too many keeki dayo it's too dangerous to go further!
Joooolien
>>109479527personally I add "plump" to my gens to make them more to my liking
>>109478859https://huggingface.co/kohya-ss/Anima-LLLite
>>109476680>>109476731>>109476688thanks I'll give them a go >>109478248yeah see >>576270912>>576203329
oguri feet
>>109479527screwy perspective but
>>109479527dammit I just noticed the leg error
>>109474364NAIce
>>109478435Nope, still fucked. Example this gen makes the femur absurdly long at cowboy shot range. AI still thinks the leg starts where the thigh high starts. >>>/h/8919552
>>109478016>>109477608oeoe pretty cool
>>109478523>>109478834beautiful style quality
>>109479934adorbs
>>109480594That post has notoriously long legs but I don't feel like that's that prevalent. Also that's not a zettai ryouiki.
>>109480429SD1.5 ahh slop
>>109480429>maximum detailed textures, maximum detailed shadows
>>109478847it could just be a lora but this was pretty easy
>>109481226AniStudio supports krea 2!?!?!? Fizzleqt
>>109476680>I discovered those tags using my rankerWhat's a ranker anon
>>109476659>I want some answers depending on who you areTry>tomatika, yoshikawa tatsuya, uchi vs the world, miche, yasuda akira, kawano takuji, bm94199, taro-k, pnpk 1013Uhhh that's all that comes to mind rn
>>109481299of course!>>109481317nice!
>>109481772Wait, why the fuck does Saba have a halo
>>109481535>One Piece anime style that doesn't look like crapIs this on Krea?I've been disappointed in Anima's lack of anime loras, specially compared to illustrious
>>109481783You didn't see anything.
>>109456499i'm not sure if you're thinking of the same person as my gpu never broke but i did have issues with my machine that prevented me from genning which i have fixed recently. i don't think it can run Anima, at least not comfortably. i've grown to hate waiting 30 seconds for a single 1mp sdxl gen. aside from nai being the most convenient way to gen atm i use it cuz i like it, even though it really annoys me that they don't offer a good way to upscale or to take images beyond 2mpi didn't get the chance to play with Anima but i've seen some really good stuff from it and can't wait to use it once i get a better rig. especially can't wait to train loras
>>109476977that's what happens when you use a 1b param shitbase designed for factory robots instead of an actual good base like krea designed for art
>>109481819>Is this on Krea?yes>'ve been disappointed in Anima's lack of anime loras, specially compared to illustriousthis is just the base model. it lacks in booru fanart styles and nsfw past booba but for anything at least moderately popular it knows it pretty well. It mixes styles better than anima too. still needs more loras for nsfw artists though but what it really needs is a finetune
>>109481994>this is just the base modelI should reiterate, this is the turbo distil, not the base model although that is available for training. it's also just a tad slower than anima but for a much bigger model
>>109482020>>109481994>>109481535Liar! :(
>>109479934woops I forgot to link cross-board>>>/vg/576270912>>109481442There's something I like in this bunch, thanks.
Why did nobody tell me that you need to add a trigger word when making a Krea2 Lora otherwise it doesn't trigger?
>>109480429new models just don't reach this level of slop anymore
>>109482413Might enjoy Yusuke Nakano too
>>109482493that one's great, fixes the faces and adds some life to the expressions
>>109482126>>109482484Do you use any UI for NAI?
>>109482394I don't know what else to tell you other than I am too lazy to update all the docs
>>109482552used to use the A1111 extension but not at the moment
>>109482582that is a cool looking ui
>>109482598ty
There is a comfy node for Anima's artists mixing. Is it possible to vibecode it into a Forge Neo extension?
>>109482463Seriously? I downloaded a bunch of loras for Krea2 and most of them worked without trigger words.
Love ani
Protecc Ani
>>109482694Weird, I left all the settings at default on the Civitai trainer, but made sure all the descriptions were correct (which took me almost an hour).I'm trying on both Neo and ComfyUI and it doesn't seem to apply the style.
>>109482677Which one? They're sloppy, I tested it in ComfyUI and I ended up sticking with Prompt Schedule, which actually works in Forge.
Say if there’s a Minimax anon in this thread, have you tried using an IRL video scene as a reference, then using an anime character as a reference image?Does it actually do the style transfer or it just fully replaces with the same background?
>>109482735I was thinking about this one https://github.com/An1X3R/Anima-Artist-Mixer
>>109482758i believe without specifying in the prompt itll just place the anime character in the real background.
>>109482758just ask the /wsg/ thread. there is an anon with some RWBY example he showed
>>109482598check this man's HDD
>>109482795Interesting, I haven't tried that one anon, but I'd guess it could be vibecoded for Forge, they probably use a different lexicon or naming to denote things. Good luck with that!
