Selfie EditionPrevious Thread: >>>>8689703>LOCAL UIreForge: https://github.com/Panchovix/stable-diffusion-webui-reForgeComfy: https://github.com/comfyanonymous/ComfyUI>RESOURCESLora Training: (NEW) https://rentry.org/yahbgtr | https://rentry.org/Copier_Method_Notes | https://mega.nz/folder/VxYFhAYb#FQZn8iz_SxWV3x1BBaJGbw | https://rentry.org/59xed3 | https://github.com/derrian-distro/LoRA_Easy_Training_Scripts | https://github.com/bmaltais/kohya_ss | https://github.com/Nerogar/OneTrainerControlNet: https://rentry.org/dummycontrolnet Inpaint: https://rentry.org/comfyui_guide_1girl#up-to-date-comfyui-guide-for-1girl-and-beyond | https://rentry.org/fluffscaler-inpaint | https://files.catbox.moe/fbzsxb.jpg4chanX Catbox/NAI prompt userscript: https://rentry.org/hdgcb>FAQ>What is model is everyone using nowdays?Any of those - NoobAI-XL (NAI-XL) (base model, BEST FOR LORA BAKING) | https://civitai.com/models/833294/noobai-xl-nai-xl FOR DOING GENS:- 102d custom| https://civitai.com/models/1201815?modelVersionId=1491533- 291h | https://civitai.com/models/1301670/291h- r3mix | https://civitai.com/models/1347947/r3mix-illu01nai-v-pred>How do I do XGiven how the models work and are trained, you should always look for a corresponding tag in these places- For anime -https://danbooru.donmai.us/wiki_pages/tag_groups | https://danbooru.donmai.us/related_tag See tags in action: https://tagexplorer.github.io/#/- For furry-https://e621.net/tags | https://e621.net/pools>What I want to do doesn't have any tag or tag related to thatTime to train a lora (see the resources section)>How do I add/remove/fix thingsInpaint (see the resources)>I want to do X interaction and/or put Y amount of characters at the same timeYou can try to prompt accordingly for it but you will end up using controlnet and inpaint (see the resources again)OP Template/Logo: https://rentry.org/hgg-op/edit | https://files.catbox.moe/om5a99.png
shitmixes shouldn't be in the OP. Period.Fuck off with that shit.
>>8693977show your base vpred 1.0 gen with metadata
>>8693979https://litter.catbox.moe/k69jgwk70kpas11z.png
>>8693982sorry, but this is not exactly helpful:upscale metadata (useless for base gen settings)unidentifiable lora at 1.0 strength (might as well just be that slop stabilizer for all i know)
>>8694003https://files.catbox.moe/hazb8u.png>unidentifiable lora at 1.0 strength (might as well just be that slop stabilizer for all i know)A single lora is a single lora. And all you need to know is that the dataset for it is from someone who is an inspiration to us all.
This one is really silly considering how the background is done, probably I should have dropped the sketchbook thing
There, 3 images
cant find good settings to inpaint a detailed pussy on r3mix...
>>8693972why do you keep rebaking this crap...
>>8694051maybe don't use r3mix thenanyway once you have enough denoise for the change it's all art style
>>8694079I got to a decent result I think, also did a lot of inpainting on the character and this is the final. I'm pretty satisfied with the results overall but I will definitely play around with more models, enough drawing Makima for now though.
>>8694090
>>8693972Why did you remove the ez-scripts fork, retard? nobody is using base anymore
>>8694184He removed most of the useful resources and replaced them with a shitty FAQ
>>8694184Because the shitty cumfy tutorial that recommends Illustrious 1.1 and WAI is more important.
>>8694184because he's a /g/ vermin that doesn't know the actual resources and wanted to make an epic op with shitmixes in it, what else?
>>8694242Thankfully it's just the split thread so no harm done.
Really wonder what anti-split schizos expect to happen if we returned to the main thread
at this point this thread gets more shitflinging by said gverminbot anyway so maybe don't come back to hdg?
>>8694250Clearly late '23/early '24 /hdg/ would magically return. Everyone would just stop acting like assholes, parroting forced memes, spamming non /h/ gens, and freely share metadata.
>late23/early24 hdg>goodWhat a newfaggot????
What dataset editors exist out there, anons? What do you use?
excuse me, groidie and birdschizo was actually peak
>>8694287waifuset cause the dev is cute
>>8694287Like tags? The webui extension for deduping tags, taggui for large datasets, and my scripts for automatic cleaning.
>>8694292based>>8694294>tagguiIsn't that a little old? I think there were more recent UIs
>>8694300i guess? it just works though
oh wait i just realized that taggui also has a tag dedup thing lolcan trash that shitty extension now
new VAEhttps://www.reddit.com/r/StableDiffusion/comments/1msdho7/sdxl_anime_vae_update_they_are_the_same_picture/https://huggingface.co/Anzhc/Anzhcs-VAEs/tree/main
>>8694324Why is it so blurry?
>>8694324ground truth where sir
>>8694329here *spreads legs*
>>8694324marginal improvement on colors and detail resolution, particularly things like pubes
Golden noise noob finetune will save local.
>>8694184>Why did you remove the ez-scripts forkWhich one?
>>8694418https://github.com/67372a/LoRA_Easy_Training_Scriptsand this https://github.com/67372a/sd-scripts
>>8694420None of those were on the "original" OP rentry but yeah whatever, I'll make space for them>https://rentry.org/hgg-op/edit
Is there any way to freeze the te in ezscripts after some portion of training or is the only way to save states and manually keep track of lrs and training time
>>8694428none that I know of, only save state then resume with different settings
>>8694324do those still meltie when inpainting?
>>8694499settings issue
>>8694324well i tried it and it only made my oekaki meltier, not sharper
>>8694499They shouldn't melt even with the default VAE if you aren't retarded
left is the "better" vae and yeah it just looks like someone ran animemeltie 6b over the lines
oekuckie bros...
There a guide for the tiled upscale meme?
>>8694025>>8694027Nice. Is this an img2img or did you prompt for the notebook and such?
>>8694263>Clearly late '23/early '24 /hdg/ would magically return.I guess you're a newfag like me. Still those were the days. Even when I showed up here in Nov '23 the oldfags claimed /hdg/ was already shitty and that earlier in the year or even 2022 was far far better. It would be nice if we could just get the '24 era back at least.
>>8694324they're the same picture
>people unironically nostalgic for a graveyard era generalwowie
>>8694526oh wtf i tried itdo people genuinely use that garbage?
>>8694526The old guide was debunked and shitters never explain but I can post my settings.>NoobXLControlnet_eps tile model>tile resample preprocessor>control strength 1>start step 0; end step 0.8>balancedgogogogo (at 1.5x resolution). End step functions as a "fixer" as in the lower the end step the more radically everything in the image can change (and give you bad anatomy). The higher your end step, the less it changes regardless of your denoise. You'll need to play around with end step a lot, especially with loras.
>>8694559denoise? also, straight img2img? half of these jeety guides used ultimatesd for it
>>8694526I use>1 denoise, 0.35 strength, 0.75 endFor the first pass>0.85 denoise, 0.85 strength, 0.85 end For the second pass, changing around things a little depending on the style. The first pass is rather aggressive, but it's generally beneficial in fixing the base gen's small mistakes, you can increase the CN strength and end step if your base gen is good already.
>>8694538Pure prompt
>>8694559That sounds more like upscaling with tile controlnet, not tiled upscaling.
>>8694563I use 0.4 but use whatever you want. At 0.8 end step I haven't noticed much improvement with 0.95 denoise, just a lot of time wasting.>>8694568I don't know what you're referring to then. Ultimate SD upscale is shit.
>>8694564Also as an added note, I noticed that generally, lower CN strength values result in thinner linework and a higer level of noise in the pic that may get interpreted by the model as more detail or just left there to be an eyesore depending on the artists and loras.
>>8694565box?
>>8694571>https://files.catbox.moe/l1lfbr.png
yeaaaah nah i can't see any discernible difference with either guy's setups compared to straight img2imgobviously it's more stable with controlnet at silly denoise strenghts but it predictably doesn't upscale any sharper
>>8694578>obviously it's more stable with controlnetYeah that's how I would describe it. You're in full control and can shift end step and denoise around depending on your needs but quality wise idk. For some artists it's better but it feels like placebo. Regular img2img tends to give a lot of hallucinations and other bullshit so I just prefer CN and adjust the numbers as needed. It's very helpful when you're up at like 3500x3500 resolution.
>>8694584I kinda want to try baking some artist at 1256 or 1536 or whatever and see if it somehow improves upscaling
>>8694578>it predictably doesn't upscale any sharperyou have to raise your resolution even more my friend
>>8694591that doesn't change the vae brobut nah i tried two different ones and it's literally the same line clarity
>>8694569nta, just saying they're separate things. He might not know what he was asking about either.Using tile controlnet while you upscale is almost the same as using xinsir union or anytest, only difference is accuracy and which parts the CN focuses on. Tiled upscaling is meant as vramlet cope and for extreme resolutions. Breaks the image up into small tiles and upscales them one by one. So you won't crash on out-of-memory or start swapping to disk even if you go like 8K res, it'll just take a long time.
>>8694590just use the 1536 lora extract
>>86945261.5x base gen kohya deep shrink2x upscale, w/ realplksr https://github.com/Phhofm/modelsdownscale to 5MP total, denoise last 7 of 30 steps w/ low strength promax controlnet, using base gen as control imagedownscale that to whatever fits in a 4MP png for export to 4chan
>>8694593>that doesn't change the vae brothe entire point of upscaling to insane resolutions is to overcome the shitty vae
our helpful chatgptbro :>
>>8694597>realplksrqrd? been using animesharp for 2 years already, anything else was either too shitty or too slow (dat upscalers)
>>8694593The VAE issue is just requiring high resolutions to produce proper details, and requiring high resolution images (or close-ups) in training to learn those details. The way upscaling/adetailer/inpainting fixes eyes and other small details is by increasing resolution. That's all it takes, at that point the VAE doesn't limit you in any way.
rmao
>>8694526use xinsir(from what little I've tested of euge's version it seems slightly weaker and you want the strongest controlnet so you can use weaker settings), .35~.45 weight(adjusting up when you need things to stick better), .45~ end step(this is long enough that it'll, largely, set the actual image but short enough that it'll be less constrained when filling in details). Denoise .65-.8, higher is preferable but again, if you're having issues getting it to correctly interpret things you can lower it to force it through.ultimately though this is all something you have to actually play around with and use a lot to get a feel for since all the ranges can be varied for different reasons and different purposes.
>my favorite style from latest bake is 7 imageswaow>>8694596I tried that, doesn't really look that good either.I'll give the bake a try, obviously won't be good for base genning but it might do some weird shit to help upscaling, it'll take like 10 minutes anyway
>>8694595>Breaks the image up into small tiles and upscales them one by oneIsn't this what multidiffusion does?
>>8694601realplksr is much sharper than animesharp, much faster than DAT, i generally prefer the variants that have minimal noise handling but even the "real" one is fairly gentle on stomping detail
>>8694610>much sharper than animesharpthat isnt necessarily goodI'll give it a try though thanks
>>8694575Oh I didn't know crosshatching was different from hatching (texture). Ok thank you.
>>8694608Yes, multidiffusion is one tool for tiled upscaling. Mixture of diffusers and ultimate SD upscale are older ones.
>>8694613no significant ringing from what i can tell, the minimal noise handling versions are also better on avoiding bandingand the difference in clarity on final gen is very significant coming from animesharp
>>8694617Which model in that long list?
>>8694324it does sharper lines but also is giving me an insane banding effect (which i don't like)
>>8694610"real" ones indeed seem to be really sharp and don't look too bad from the first glance, but "layernorm_gan" doesn't feel sharper than 4xanimesharp to mei'll find a use for the real ones though
thanks again to the anon that recommended that dataset editor last thread. Idk why Tagger is so fucking slow but this one is super fast.
ok no the 1256 lora didn't help one bitdamn you vae
>>8694644What dataset editor?
>>8694704he's probably referring to >>8690197i also added some other uis i managed to find into this list https://rentry.org/yahbgtr#editing, haven't checked every one though yet
kohaku just released another future SOTA model waiting for a bottomless angel to finetunehttps://huggingface.co/KBlueLeaf/HDM-xut-340M-anime
>>8694736how gud is the vae
>>8694736tbdesu that actually seems like it could be really interesting.>>8694738https://huggingface.co/KBlueLeaf/EQ-SDXL-VAE>it's an SDXL VAEDOA
>>8694741kek my shitposting wins again
>>8694741>340M>interestingmight as well hope that your toaster burns the image of a hot anime girl in your breakfast by accident
>>8694736>1024px>4chinz vaejust put me down already bros
>>8694741>it's an SDXL VAEit's not compatible with sdxl itself doethat said that model is tiny, you prob won't get anything useful out of it
>>8694736>webp dataset againbut why?
>>8694746it's interesting because lighter-weight models ARE interesting. Not everything needs to be ultra-HD 4k BLU TEH REI tier resolutions.>>8694748still the SDXL VAE(finetuned).also this seems even more dumb since smaller models would actually benefit more from a high channel VAE than larger models.
>>8694752more parameters = better
>>8694753yeah how'd that turn out for flux and chroma?
>>8694747Well he specifically said he wanted poorfags to be able to use it. I've been telling you guys that models should be made high quality regardless of whether or not people can run it. If it takes 5mins to get your awesome 2000x2000 base res gen then count your blessings and download more vram. You'll enjoy kino one day. But if we keep chasing the "can I run it on my 6gb laptop??" retards then we're fucked.
>>8694751Unless they compress it below 95 quality, what exactly is the problem with that? Storing all that shit in png is just a waste of space
>>8694752>Not everything needs to be ultra-HD 4k BLU TEH REI tier resolutionsYes they do, sweaty. If you want to gen in low res on that model you still can but if we do it my way, everyone benefits.
>>8694755flux was sota when it came out and chroma was made by some retarded furry
>>8694755>>8694762did chroma flop?
>>8694755poorfag animebros not being able to tune it is not a measure of success or not desu, it was the same shit of waiting 1.5 years for a half component sdxl tune (which was also derided at the time lol)the plebbitors have been using flux for a while
>>8694758you are training an image generation model, if you give it the same pattern for EVERY SINGLE IMAGE, i dont doubt its gonna start including it in every single of your generated images by defaultmore storage space for the data is gonna be maybe like 200-400$, and you are already spending thousands on training itself
>>8694758Why gimp yourself unnecessarily is all. Just do it the right way for the peace of mind, you only need to succeed once.
>>8694755Flux is literally local dalle-3 m8. It was only superseded by sora.
>>8694766Yeah I understand that, but you can use lossless with webp and it would still take less space than png. And I'm not really sure if 95-99 quality compression "pattern" would get picked up by a 4ch vae at all. Noob's issue iirc was that they used like 90 quality, although I might be mistaken
>>8694762no, people wanted to pretend it was when it wasn't. And then died down when people realized it was untuneable garbage.also """"""sota"""""" means jack fucking shit in this space.>>8694765>n-no you're just poorfags! those are the problem!oh fuck off you retard.and I'd be interested in this shit because a model trained at 512x512 with a 16ch VAE on a pure chinese cartoon dataset with, ideally, updated *booru tags would ACTUALLY BE INTERESTING. I couldn't give a fuck about "muh tekkknology!" dickwaggling bullshit. I want to actually make art with this shit.
>>8694770yes i dont have anything against lossless webp, but there is literally no reason to use lossy webp other than saving a few bucks on storage, just why risk itcompressing jpeg images even further and compounding whatever artifacts they already had instead of actually trying to mitigate them is insanely retardedthe dataset kohaku is using is also 90, and i assume every trainer using webp shit is gonna use it as well
>>8694773>no, people wanted to pretend it was when it wasn't.It was the best generalist model for nearly a year and widely used for everything that wasn't porn.
>>8694773We already have a 16ch VAE 512x512 model though, it's sdxl. If you downscale your 1024x1024 output images to 512x512, you'll get 16 channels per 8x8 square instead of 4.
>>8694776too bad we're on a porn board then, huh.
>>8694770>lossless with webpjust use jpeg xl, it's even better
>>8694773>no, people wanted to pretend it was when it wasn't. And then died down when people realized it was untuneable garbage.When it comes to prompt adherence it's as good or better than dalle-3, its main issue is awful styles without loras.
>>8694779>>8694765
why are the retards from /g/ here?FLUX is irrelevant. It has always been irrelevant. Even in chinese made AI advertisements they don't use FLUX. They use SD1.5 for that shit. SD1.5 is unironically significantly more relevant to the world at large than FLUX has ever been.
>>8694770>Noob's issuei think the main problem was that the gpu servers were geographically located in china and they couldn't distribute the dataset super well (i think they were provided the cluster but not funds)
ugghghhh sdxl is so bad it can't do anime and was trained on 3dpd(it's totally not deriding it because we are too retarded and poor to tune it lolll)oh pony and noob? i looove sdxll!!!
>>8694786Nobody cares about flux but it's simply a fact that models with more parameters are better if you can run them. Retards that want a low para model are invited to download jordach's magnum opus cascade 1B
>>8694786You are very retarded. You're not discussing model quality but difficulty in usage. In fact, I know you're not stupid, you're just being dishonest by pretending you don't understand something so simple. Flux is too difficult to run and because of that (and because of the lack of competent local bakers) there was no finetune. If there was we would be eating good.
>>8694770You are mistaked, that was Pony with the jpegs
>>8694765>the plebbitors have been using flux for a whileand they have no taste, honestly.The flux face is a real thing. Plastic skin. Just general lack of art style support too, not just anime, it can't do shit without LoRAs.The only good thing Flux has is prompt adherence. But then you have to write a novella in your prompt to get anything good.>dalle-3I hate flux but it's actually better than dalle-3 and the only time you should compare something to dalle-3 is if you want to insult the model you're comparing it to. Oversaturated garbage.Between that and 4o piss filter the image generation field is not, in fact, in constant evolution, it's devolving.
>>8694799Flux should be compared to SD1.5 and SDXL, until we get an anime finetune the face skin etc are irrelevant because they'll change. Same with natural language prompts.It's a good base when it comes to prompt comprehension, knowledge of everyday objects, sfw poses, etc. But wihout an anime finetune it's only usable for setting up /e/ scenes for img2img. A shame Chroma was only pretending to train on anime, it's apparently pretty good for photorealistic furries.
>>8694796>Flux is too difficult to run and because of that (and because of the lack of competent local bakers) there was no finetuneIf you think that's the reason for FLUX's lack of a finetune then I've got a bridge to sell you.People wasted a lot of money trying to finetune it only to find out that it just cannot "learn" more than like, two things. You want to teach it a chinese cartoon art style? Cool! It can do that! Wait you want to teach it multiple DIFFERENT chinese cartoon art styles? Woah woah woah, hold on there buddy. That aint gonna fly. How about you teach it to draw a really cool ferrari instead of those other art styles, yeah? Wait. Stop. What do you mean you ALSO want to teach it how to do a Lotus Elise? etc. Shit is hardstuck on what it can learn and how it will learn it.
today i will remind them
>>8694798I remember there were some melties about noob using compressed webp dataset. https://huggingface.co/Laxhar/noobai-XL-1.0https://huggingface.co/datasets/NebulaeWis/e621-2024-webp-4Mpixel?not-for-all-audiences=truequick glance through noob info and it seems like noob used 90 quality webp for the e621 part at least. Pony's jpeg artifact issue was immediately obvious, gens were consistently noisy, and he did something like recompressing it twice or smth
>literally zero narrative change between every "too poor to tune" bake lol
>>8694805SDXL had dozens of random normie finetunes fairly fast.FLUX and chroma have both had nothing but failed attempts.Also I'm not sure what your point is when at that time there was literally nothing you could do with SDXL that was relevant to this board.
jej
>>8694806they used it for danbooru as well
>>8694809I mean other than the boobs, he's not wrong. Turns out slightly more parameters is all you need.
>>8694808there were couple of very annoying sai/cumrag shills pushing for the garbage sdxl shittunes, posts shitting on sdxl are likely a response to those shills.
>>8694809In the end, people did use SDXL as bigger 1.5. Nobody cares about refiner and 2 TE bullshit that were part of sdxl
After testing the finetuning meme for a while I came to the conclusion that lora training is superior as long as you're constrained to <=24gb vram. The quality loss you get from shit like 8bit optimizers, full bf16 or stochastic rounding is just too severe.
>>8694815>as long as you're constrained to <=24gb vramWhat if you aren't?
>>8694808it took quite a long time to start to get usable tunes for sdxl, i think ponyv6 at the time of release being considered the second coming of christ is pretty telling
>>8694814>Nobody cares about refiner and 2 TE bullshit that were part of sdxlOh yeah, whatever happened to that shit? Is anyone still using that?
>>8694817you could probably make some very nice lora extracts if you rented compute
>>8694804because it's the distilled version
>>8694821cue 50 different snake-oil salesmen popping up claiming they are totally undistilling it and it'll work this time I swear!>>8694818oh should also probably note that when SDXL came out a lot of people were still kicking with 10-series GPU's. By the time PDXL came out people had finally migrated to 30 series since it actually offered a tangible upgrade in cost/performance.Meanwhile it's 2025 and brick and mortar stores are still stocking 12GB 3060's.
Can you guys post some gens between your model rambling so it doesn't look like I am the only one doing that, don't get me wrong, keep doing it just attach some gens from time to timethanks
There won't be any breakthrough in image gen until compute gets cheap enough that a sufficently dedicated hobbyist can train/finetune a model without cutting corners. You need to throw shit at the wall until something sticks to make any real progress besides scoring high on hermetic KPI dashboards, and at the current prices that's simply not possible for anyone but corporations which can't possibly give less of a shit about image gen, let alone about anime image gen.
>>8694832Why would I post pictures of my wife for your to leer at?
>>8694130different style but very nice I like it
>>8694834which is why a smaller model is INTERESTING. It doesn't need to be something super crazy amazing displace everything sort of thing but at least people will actually be able to do something with it and that makes it INTERESTING. I think the approach here is dumb and bad since it's still an SDXL VAE but resolution is honestly a corner I'm "okay" with cutting, especially if you can use the param drop to instead utilize a heavier VAE.I went back and trained some SD1.5 loras a month ago for shiggles(since naiv2 was open sourced) and man, I REALLY fucking miss how fast that training was.>>8694832why do you want to get me banned
>>8694832we've been through this billion times alreadypeople either don't gen extremely narrow "/h/-related" pics (penis in hag vagina, anything else is off-topic) or don't bother with doing "effort" gens that often, so the thread is just filled with text for the most partrule trolls won, so those are the threads you get
>>8694836thisposting gens is virtual cuckolding
>>8694846banned for what lil bud>>8694848I am not expecting complete effort /h/ approved gens, anything goes really as long as it's not actually /g/ shit or something like that
>>8694815full bf16 is supposed to increase quality isn't itanyway 1:1 on loras at least gave me identical to better results on 8bit and full bf16
>>8694846>use the param drop to instead utilize a heavier VAE.That's not how it works. You need more params to store all the additional details the vae pic up during training
>>8694832>>8694852T-Take my peace offering...
>>8694861/h/?
>>8694861Oh uhm, it needs some work...! but thanks for posting
>>8694865Bro... I didn't know pussy juice was /e/>>8694867I-I did work on it...
>>8694868>I-I did work on it...I, hmm, recommend zooming in and out a few times next time! gambare anonie!
>>8694849If you post your wife, you're a cuck.if you post another girl you're cheating on your wifeOnly by not posting images you can truly love her
>>8694873based vanillachad
>>8694873I don't post my wife doe, I post yours.
>used 3060 12GB's are still selling for $200+ USD on ebaythis is hell
>>8694881My wife is "skinny, slim legs, small breasts" fyi.
>>8694855>full bf16 is supposed to increase quality isn't itno lol, mixed is always preferable unless you're some sheik and can train in fp32
>>8694832here you go sir
>>8694912Is this magion?
>prompt for anon's wife>only comes up as blonde hair blue eyesintriguing
>>8694912very based
>>8694916it's godiva ghoul, (suenari \(peace\):0.6), (itowo \(itowc\):0.8), hatching \(texture\), with this lora at 0.4 https://civitai.com/models/1127584?modelVersionId=1528491
>>8694855>full bf16 is supposed to increase quality isn't itnot really. usually you are doing mixed precision training (sdxl itself was trained this way) which means that the model is stored in fp16/bf16 and the gradients are fp32. full bf16 by itself converges noticeably slower due to precision loss in gradients..
>>8694815what did you try doing?
>>8694924Training a style. tried adafactor, lion and paged adamw8bit with a bunch of different settings. results weren't terrible and finetuning was always able to pick up some parts of the style that i never got with loras, but the overall result still felt kinda lacking in comparison.
>>8694917How is that small breasts though?¿ Also hourglass unerotic.
can someone give me a higher definition of this pic? A bit more mature girl, with small boobs and a large ass please. Thanks, anons
>>8694932>but the overall result still felt kinda lacking in comparison.that's very vague, can you elaborate more on that?
>>8694943adding another just to contriboot to the thread, but wrong pose
>>8694943>>>/trash/slop has requests and billion sloppers hungry to fulfill them.
>>8694944Anatomy sucked, backgrounds were incoherent, trouble with small details in general
>>8694937because I wanted to actually post it to the thread without being bannedhttps://files.catbox.moe/fa5ric.png
>>8694949just from the way you describe it, to me it seems like it's more of a dataset issue which is somehow less noticeable when training a lora (likely because lora training has much less expressive power)>finetuning was always able to pick up some parts of the style that i never got with loras>trouble with small details in generalyou are borderline contradicting yourself there.
>>8694947Tried but no luck? Maybe I'll follow your advice and try again anyway
>one of the stronger oemeltie styles just doesn't train inhm
>>8694958>you are borderline contradicting yourself there.finetuning captured the vibes better or perhaps the sublte meltiness just helped with the more painterly aspects of the style.>seems like it's more of a dataset issueI tried different datasets and even included some face/hand crops but i didn't make a difference. I blame it on lacking precision
>>8694970Uuoohh contrast
>>8694970I fucking hate when I put "dark skin, interracial" in negs and bro comes out grey or yellow or purple or whatever the fuck except white. (I realize you probably prompted it this way)
>>8694976No, he's just built like that. I don't really care about skin color for the dude to be honest, I kinda cancel it out when looking at a picture especially if it's POV.
>>8694972>sublte meltinesswell, if you put it that way...i think i can relate. have you tried turning off/reducing layer strength when applying a lora? would be funny if this meltiness is somehow related to adaptiong to a tiny melty vae dataset better or accidentally frying some layers.
>>8694987No, but I guess I can try that tomorrow
>>8694992please post about it if you find something interesting.actually that webp discussion before made me think...what would happen if you also compress your dataset using 90% webp? could it be noob didn't learn the tiniest details well due to the entire dataset being compressed?
>>8694921Thank you!
>people actually schizoing out about "compression patterns"i've been trying to bake in severely heavy jpeg noise into some of my loras and that didn't do anything
>>8695018Ponyv6 (actual pony, not autismmix) reproduced jpg noise very well.
>>8694943here>>8694976>he cares more about the cock than the girl with come on her facekek>>8695018This. Even when I bake eencya it smoothed out the pixels.
>>8695021What the. I guess that was the wrong one.
>>8694832
>>8695018there's a good chance it's being filtered out by your negatives and/or banked somewhere in the model that recognizes it as something extremely specific and needs you to actually prompt for it.illustrious models have a really, really bad habit of doing that to "traits" (whether you tag them in your dataset or not) like sketch, painting \(medium\), etc.
>>8695021>>>he cares more about the cock than the girl with come on her faceYeah that's because I don't just do "the cock".https://files.catbox.moe/96olcd.pngHaving a fucking n-word standing in the middle of a gen is revolting and not my style. Send them back to their own continent.
>>8695026>negativesshiggydiggy
>gay pedo meltie
>>8695027American website so this is my continent, sir. (You) on the other hand clearly don't belong since you need to catbox your own pics, unlike me :^).
kiss no erotic...
>>8695035I don't belong together with cotton pickers? I know. That's why I don't live in a place infested with subhumans. The internet is too global though, that's yet to be solved.
>>8695030>using an overfit shitmix>AND falling for dumb snakeoillmao
>>8695042Exactly! So haven't you gone back to /b/? There's no coons there, only your kind do that shit.
this guy is trying so hard to fit in wtfthat's like 2013 edgelord talk
>>8695046Same reason you haven't gone back to Africa.
erm that just happened
>don't know a bunch of the styles enough to know if the very slight stylistic changes after resizing are just model melt or the actual style
>>8695050To fuck thick sluts that give it up for free? Huh?
ok now it's samefag
>>86950674chan is 1 person btw
>>8695068have you seen the archive reposts in some of the boards? you're so right
>>8695066Well. Good for you and that's your business.I was talking about my _gens_ though before you or some other bro said "I shouldn't care". Well I care, I don't gen n-words. That's my business.
why is lilnigger afraid to say it on 4chinz?
>>8695074Good to finally see some gens. Cheers.>>8695077Cuz he thinks I'm going to banish him back to /b/.
>>8695077Having to wait 120 seconds is inconvenient.
>not a single comic in dataset>artist doesn't even have any on danbooru>gens a comic AIEEEE
>>8695080Don't use close-up
lilbro thinks he's a mind reader
>>8695090>>8695092perchance?
Does anyone know what happened to Sankaku?Why is it always showing me the mobile site now?
>>8695095i couldn't even load it a few days back so i think they're fucking with the site
>>8695085I'm right.
>>8694324...it looks worse? And he even has the gall to ask for money.
>>8695109many such cases
But when I (ME) post Marnie...
>>8695118I liked your gen so much I wanted to do my own. Amazing Marnie. Sorry you have to wait 3 days every time this reddit sub goes retarded.
>>8695120>following the rules is redditYou need to go back.
>>8695120To be fair I was quite susprised, I think she looked old enough and nothing like a loli at all but I guess the style didn't help my argument that much
>>8695115
>>8695123She was fine, the state of this website is not.
Inpaint detailing practice>>8695124what model?
>>8695150102d
>>8695194Will you start screaming BIBIICIII too?
>>8695194
*ahem* BIBICIII
nose
painterly NEXT
I've been fucking around with the new-ish video gen in comfyui and despite working well this morning it's totally shitting the bed now (ex: dipping outside of allocated vram and slowing to a crawl for seemingly no reason when it was fine earlier). Anyone know what might be causing this?
Does anyone have the metadata for SparklingSenpai's gens?
geg
>>8695236$25
>thread split to try and get away from shitposters>but because this thread was a lot more active earlier the shitposters are just here now and it's indistinguishable from standard hdg garbagewaowwho could have possibly predicted this would happen
retard
>Man I like how sharp this old gen looks, what were my settings for base gen on that time>the same sampler and scheduler combo that I am currently using but with 6 of CFGOh...
placebo
>>8695270Why is everything but her eye blurry?
>>8695270based orc god
>>8695018oh well lets just compress our dataset twice with 90 jpeg then, surely nothing bad could happen
>>8695271Crop factor to low on the inpaint, please understand.
dumb faggot
me?
Hello, can I gen a smarter, more detailed model?I'm kinda tired of getting dolphins every time I prompt for dolphin shorts. And eyes being melted when genning something bigger than an upper body. And the model not being able to understand characters with asymmetric features, even while using LoRAs. And not being able to gen text. And... Well it's a long list. But you get it.
>>8695324better learn to inpaint bwo, none of those are problems at all
>prompt 1girl>CLIP now pulls from every tag containing the letters g,i,r, and lFUCKING CLIP
>>8695327What if I don't have a top tier GPU? What if I just wanna hit the generate button while jerking off?
>>8695333haha well NovelAI only 25 a month my friend!
>>8695333Then why are you complaining? lmao, are you one of those that want 1-click everything?
did someone say $25?
only 25 a month for SOTA anime and hentai model? sign me up!
>lol! you wanted your software to work?
>prompt doggystyle>gives me a girl wearing dog themed clothesFUCKING CLIP
Yeah nah, I'm going to sleep now, you guys deal with him
why does he say this so often?
>>8695333>What if I don't have a top tier GPU?You don't need a top tier GPU, any Nvidia GPU with at least 6GB VRAM is technically enough>What if I just wanna hit the generate button while jerking offThen you should either lower your expectations for what a GPU can hallucinate out of random noise, or wait some years until the tech gets there (or doesn't who knows).
>>8695357>Then you should either lower your expectations for what a GPU can hallucinate out of random noisebut why can nai gen me some porn with one button while I jerk off and give me one gen every 6 secs but my 8gb vram gpu can't? that doesn't sounds fair
Is there any sort of inpaiting tutorial regarding NovelAI 4.5?
buy a gpu
>>8695399What sort of "tutorial" do you need? there's literally nothing easier than this shit, no 500 toggles, no mask blur or other shit you have to care about, don't really have to change the prompt even if you are doing are inpaint, there's literally nothing to write about.
>>8695426My question mostly regarding glitched fingers. Just mask them and inpaint? That's it?
>>8694976You're using noob, right? Just prompt for the dick color you want. Negs were always a cope, and they somehow work even even worse on illustrious stuff than they used to on Pony or 1.5.
>sd1.5>stacking loras>spend hours adjusting weights to get right output.>gens keep varying massively >change "huge ass" to "massive ass">everything is fixed.Wtf am i doing wrong.
What's with this sudden splurge of shitposting?
i noticed.
>>8695430That's what I usually do. I didn't really have much issue with hands on 4.5 although I guess that depends on the artist tags.
>>8695436i don't think anons should be allowed to post gens in this thread that don't have "galaxy penis" somewhere in the prompt
>>8695443>Wtf am i doing wrong.Using 1.5
>>8695445>thread gets more active because there aren't shitposters to ruin discussion>thread activity attracts shitposters>shitposters frive regular posters away>shitposters leave for more active threads>repeat from step 1
>>8695462God damn it.I downloaded so many lora's for 1.5.I'm just going to stick to Illust.
>>8695021>>8695023Man, thanks but as I was saving your previous gen you deleted it. The second is a bit different (6 toes, creampie). I literally was ok with the previous, that was cleaner, like this >>8695150that sadly has no shown feet. But thanks anyway!
>>8695464sounds reasonable but i don't see how this>thread activity attracts shitposterswould work, why would they shitpost in a different thread?also what the fuck is thishttps://rentry.org/lora_train
>>8693972any online image gens just for doing anime-style backgrounds? That don't look like weird ass (tables with 5+ legs for example)
>>8695445I'm new here, what amounts to shitposting itt?
>>8695456kekjust showing off that you can do whatever you want with the colors, it doesn't have to be a tag
>>8695481Try Flux on Civitai. Takes a lot of currency though cause it's a big model.
>>8695469If your machine absolutely cannot handle SDXL, you should use EasyFluff with the HLL lora instead of older 1.5 checkpoints. It's v-pred, it's got enough sex knowledge for the basic stuff, like 700 artists and some mainstream anime characters and vtubers.
>>8695483your post, because it attempts me to cite (and thus support) shitposters, disguising yourself as a newfag or an idiot or a returning oldfag to ask stupid questions, not having any respect for other people and posting slop >8695470 without even acknowledging someone might not like that, pointless shitting on other anons and trying to act arrogant, framing loli as "pedo", posting western-style images that should be on /aco/, "naiggers must die" and "it's just $25" kind of interaction, meaningless short posts without a reply or posts using gen alpha zoomer discord slang or any attempts to turn the discussion into a chat by any other means, etc
>>8695500I probably can handle xl.Its just i saw so many lora's for 1.5 i thought it was the way. Or i guess i should say I have a lora i wanted to use and it was for 1.5 so i've just been using that.
>>8695511Thank you for the explanation.
>>8695517Wrong file.
>>8695516The reason there are so many loras for 1.5 is that you can train one in 5 minutes and that you need a lora for pretty much anything.
>>8695523wow Did not know that.Yea these gens weren't taking long at all to make even at 512 hirez x2 so i kinda started to think maybe i should go back to Ill. Is ill good for semi realism ? Maybe I'm retarded but trying to do short stack is so annoying and im running local.
>clueless newfag bit again
what did he bite into bwo?
>>8695270Never enough fischl
i mean... 25 dollars is a pretty fair price all things considered
less and less are saying this.
>>8695537That advice was aimed at producing anime. You might be in the wrong thread, try >>8944432 or >>8925465 depending on the degree of realism. They generally use models based on illustrious or noobAI but pushed towards a more realistic look.
>>8695583>>>/aco/8944432 >>>/aco/8925465
>>8695303>ff14 autismNice. I made a lora of my character but I still haven't plapped her yet.
>>8695333Unironically that was the point of novelai. If you can't run local for free* then you rent their gpus and use their model. If you wish to do local then you need to put a little more effort in that set up wildcards and generate, but not *that* much effort. Most noob models aren't too bad with things like prone bone, pencil skirt, or dolphin shirts like they were back on pony/autism. If you get one of those hallucinations just smudge them or paint over them in gimp and inpaint.
>>8695511>framing loli as "pedo"I love how you snuck this one in there, meanwhile my point was to tell you that loli belongs on /b/. You are still incredibly retarded for failing to comprehend such simple rules but alas, this is your elaborate shitpost after all. Gotta sprinkle in a bit of truth to hide your lie.
>>8695488Can do glowing and other shit too I like to do ghost dicks plapping girls
>>8695470The issue was that she wasn't old enough.
>>8695481Flux by far.https://files.catbox.moe/8ld8wp.png
>>8695572Because most people can run noob, but if you can't it's better than doing 1.5.
>>8695644i didn't say loli belongs here you fuckwit>>framing loli as "pedo">retarded shitposter got triggeredchecks out so far
>>8695646wow thanks for the correction. saved. I've also got something from /trash/, in case I share with you since it's hentai still
>>8695646
>>8695644Also forgot to add, loli is any girl under 25 and not obese. Since there is often misunderstanding.
Behold, the creative power of the SOTA anime model unleashed.
>>8695669Why would you want skinny bitches? Are you a starving jeet or something? First world countries don't have food shortages and aren't scared of full-bodied women. Grow up rakesh
>>8695676What I want is for me to decide. But bending the "loli" rule every which way because somebody posts what (((you))) don't want is how you got to the current state of this shithole of a thread. (but that's what you wanted in the first place)
>>8695678It doesn't matter what I want or what you want. Moderators get to decide what loli and what isn't, and they decided that your post is, so they removed it for breaking the rules. It's rather simple, really.
>>8695680No argument here. You got to the center of the issue splendidly:)
>>8695669>loli is any girl under 25 and skinny Might as well put up those consent forms and age of charas with each post so we know they're consenting legal adults
Bros should stop bending the offtopic rule and getting me banned for bugrape and tentacles, those were always hentai
>>8695699At this point non-futa /d/ thread is less disgusting than this thread anyway and also has better moderation too. Might as well use that.
>>8695701>second post i see is quadruple amputee medusa-stone fuckingamazing thread desu
>>8695703I only said it's less disgusting than this thread, I never said it's perfect.
>>8695701That really depends on your fetish tolerance, but it is comfier for sure. And not much slower at this point.
>>8695703onahole transformation actually, he just used grey rubber in that gen
>half the posts are 832x1216 1.5 bullshit with watermarks everywhereWAOOOOOW
how come it's so dead in these threads now? Expected a bit of an influx since doing short videos local is becoming way more doable
>>8695711Thread was killed on purpose for the crime of being the best AI thread on 4chan.
>>8695710the watermarks are an issue but I really never got why people here can be such sticklers about resolution, it's ai gen bullshit anyways so it's not like you're getting more detail so much as more room for artifacts and significantly increased gen time
>>8695711Genning good shit takes skill and time, if you want a quick thread just go to one of the 500/300 slop threads for butiful 1mp sloppa
>>8695711Almost all videogens (with real anime style) are bad or way too low resolution
>>8695711/hdg/ is split in three threads right now, so obviously each only has a third of the activity. There's no going back though, fags will just keep rebaking it forever.
>>8695711Speaking of video is catbox having issues with videos lately or something? I do basically 90% loli but when I try to throw webms up there they don't actually play.
I just remembered we don't have a new logo yet
>>8695713That's true if you inpaint, or if you only gen close-ups. Otherwise you get shit like >>>/2993195 where everything's melted, her eyes, zippers, emblem on the cap, the gun, her fingers and most of the background. Upscaling is an easy fix for all of that.
>>8695721I vote for it to feature black penis and a fat hag - the match made in heaven and the two most relevant topics for /h/.
>>8695713That's wrong though. Doing tile CN upscaling or even genning at higher resolutions with something like kohya's hrfix yields much higher detail than genning at SDXL's standard resolution.
>>86957131024x1024 is 128x128 in latent space. If you are okay with that level of detail, check your eyes. (both yours and in the gens made at 1mp)
>>8695723>>8695727>some random retard says resolution doesn't matter>something only someone who has never used imagegen would say>so you decide to take it at face value and think it's genuinei got a bridge to sell you
>>8695730So you are saying all the people who post 1mp gens never used imagegen? That makes no sense, how are they posting gens then?
>>8695723>>8695727arguably the resolution of the final image doesn't matter. He may be downscaling the final result, or using a style that doesn't need fine details, or inpainting manually
>>8695731i know you're intentionally obtuse but lmao at this strawman
link to some of those gens, I want to see them but I don't want to go there
>>>/d/11353427answer my question
>>8695733What I'm saying is your rose-tinted world where someone who says resolution doesn't matter is being retarded on purpose is wishful thinking.
>>8695735>full nelson with stomach bulgemaybe /d/fags are not that bad you guys
>>8695736aren't you a fruitcake
>>8695723I just inpaint when I'm trying for something particularly solid so I guess that's why I feel that way, in my experience things tend to look worse upon upscaling but it might be something about how I've tried doing it somehow
832x1216 can be okay enough if you're not a tard genning on 1.5 and upscaling to that. Which these "people" are.
>>8695740Link to your good-looking 1mp gens, bro.
>>8695740Yeah, id depends. Think we'd have to talk about specific examples.
>>8695740>in my experience things tend to look worse upon upscalingthat's exactly why you inpaint AFTER upscaling broski
>>8695743nta but you can always downscale afterwards. Most 2K pics won't fit the filesize limit anyway, nor fit on my 1080p screen at full size.
>>8695742yeah 1.5 is pretty dogshit, I honestly don't know how I used it in retrospect with how jank it always looks
>>8695751These downscaled gens are obviously not the subject of this conversation.>nor fit on my 1080p screen at full size.4chanx dropdown toolbar -> Image Expansion -> fit width + fit height.
>>8695756oh it was the other guy, my badI mean for locally saved gens, there's no point saving them larger than I can view without zooming out.
>>8695757What if one day you buy a new monitor?
>>8695758Let's just say I'm not thinking that far ahead with my gens.
Wait what, why are we on page 5 already?
>>8695784>>8695554slip'n slidin'
>>8695752>I honestly don't know how I used it in retrospect with how jank it always looksThe mere fact of this tech existing was absolutely insane, even if it was just 1girl standing for the most part. It was hard to feel the limit of the tech when you didn't know shit and it was evolving every week, felt like you actually could do anything with the correct approach. I feel like it was Dalle-3 that disillusioned a lot of people in that regard desu.
>>8695752for me it's the tinkertrooning, same as nowi still don't think these models are acceptable in quality
>>8695436This just gives mismatching penis most of the time.
>>8695846mmm potato chips
>>8695858Multiple even
any way to prompt for push up? or tops that are otherwise holding up the breasts more than usual
>>8695879just add "skindentation" bwo
>>8695879use tags like perky breasts or bursting breasts or throw sagging breasts in the negatives
Sooooo. How do I prompt this pose on local?https://files.catbox.moe/9ecb2n.pngSaved from /b/, has meta.
>>8695896controlnet :^)
>>8695896straddling, thighjob, from behind
>>8695896girl on top,sitting on lap,looking back,pov,thigh sex,yokozuwari
>>8695896https://files.catbox.moe/stnerz.jpg1girl, loli, thigh sex, looking back at viewer, completely nude, leaning back, small breasts, lying on person, pov, 1boy, big penis,
>>8695903tits too big
>>8695903girl too small
>>8695903this looks like it's got furries in it
>>8695896Just get lucky
>rebaked the fugtrup with booru tags instead of tagger>1000x betterfucks conventional tag wisdom fr fr
>>8695633I used a friends lewd pic on img2img then did a bunch of inpainting to practice and I did learn some stuff while at it
>>8695928stale b8
bait who niggamyself and the bot's three alter egos?
>fugtrupmiss me with that tranny homo shit
>>8695959yes, but also it's AI. Forcing artists to make shit they never would make irl is the main use case.
explain again what's wrong with being a tranny homo?
>again
why are so many people posting hentai with children
>>8695972they are children so for them it's just hentai
i made a scripthttps://rentry.org/on6tfm7efirst i grokked a script to query all danbooru artists with at least 150 entriesthen i grokked a userscript to put N of them in at random into the clipboard, replacing underscores and escaping parensyou can set N, record a keyboard shortcut, and move the box around. default @match is novelai image and localhost:8188for example
>>8696009Shame the most interesting artists are below 150 pics
>>8695972Truuueeee, we should have a conversation about that!If teens are a no-no then why teens with big booba get a pass? Doesn't this promote inequality and sets unrealistic beauty standards for women?They should be equally bannable.
>>8696010you could add to replace the list with the complement, all artists under 150 entrieshttps://rentry.org/gciso7hwthis is my script for querying danbooru, maybe replacing>'search[post_count]': f'>={min_posts}',with>'search[post_count]': f'<{min_posts}',will do the trick
>>8696009that's... a wildcard?
>>8696018idk, i havent verified whether or not the artist fetch script does what it's supposed tomy guess is that line sets a filter lambda, therefore replace >= with < to filter for less than instead of greater/equal
>>8696009shame inbuilt artists are ass
Are cute Britbong anons caged and have their censoring extensions installed? :3You shouldn't be on 4chan, but gotta be safe if you are here.
Chat, do I wanna know?
>>8696022>artist:nyl2no way that does anything
>>86960314 pics on e621, unless they trained on webm thumbnails tooI'm guessing most NAIfags do not XY every token with limited gens.
at least make her anime bwo
>>8696009example usage in my genningnow that i have this big list of danbooru artists, there's probably a better way to think about what i want to grind out and automate it, but this works perfectly for when i want a dice roll here and now
>>8696053and oh, that thing i want is probably wildcards
why are mass sloppers like this
need zoomie haircut loraneed white lesbian haircut lora
https://files.catbox.moe/9lr9w7.png
how does one gen slop like this
>>8696133stable diffusion
>>8696133wai v140, big pp, pov, 2grils licky big coom
>>8695666>Satan>digits check outkek the memes write themselves. If loli is not pedo then solo male is not /y/.
>>8696143Wrinkly 110yo hands. Insides of her anus aren't in better shape either... Funny skin effect though.
>>8696143ew
>>8696143what controlnet does to mfers
>>8695798>I feel like it was Dalle-3 that disillusioned a lot of people in that regard desu.Wait what do you mean? That it made people realize how shit 1.5 was or it made people feel like local could never catch up? But yeah I feel like a lot of people here are jaded. I still marvel at the ability to see some hot obscure anime girl or gacha girl and just make a lora of her and create infinite porn. Even something as basic as pov standing doggy was a thing so rare in h-games and VNs that I'm glad I can just do it myself whenever I want. And then you add in textgen? Man idk why anyone is complaining desu.
for any schizos who are interested i've uploaded a bulk bucketing + dataset augmentation script for images with transparent backgrounds as well as script to fill transparency with fixed colors and caption them appropriatelyhttps://mega.nz/folder/VxYFhAYb#FQZn8iz_SxWV3x1BBaJGbwthe goal is basically to streamline working with koikatsu and similar programs since this makes it not matter where the subject is located in your screenshots or what resolution or aspect ratio you're working with. i'm not sure if the resolution selection ratio is as good as it could be, that's something that i should probably test that out to see how close to the sun you can fly before you start getting wonky proportions
>>8695961>2.27mb jpgDid you just slap some film grain at the end or some shit?
>>8696166I do 98 quality jpg without chroma subsampling when png doesn't fit in 4MB.
>>8696013The real answer is that mods is gods and jannies are troons who will enforce the rules whenever they feel like it. Better to err on the side of caution.
>>8696143Jokes aside, how did he do that skin texture? Guess it's the artist?
>>8696140I wish it was that easy>>8696136indeed
>>8696170Considering all the cellulite I'm 99% certain it's some 3DPD granny controlnetted.
>>8696170There's paloma piquet artist that does skin glistening like that, but you'd probably need to train a lora to capture it. This looks like main artist was laserflip though.
kina crazy how good nai is honestly
noobai? yeah
>>8696180true, novelai 4.5 (only 25 a month) is incredible!
wtf it's 2am in turkey go to bed
naijeets shilling their cowdung again
>>8696182very true thats why active threads on other boards are all novelai 4.5 users while /h/ are all local luddites
>>8696180is really good for complex cuckgens, I agree
reast obvioust samefaggus
>artist does little "when u see it" thing in every image>even tagged they still show up in half the gens as little ghosts blended with the backgroundjej
proofs?
>>8696198reminds me of dino back on NAIv3
>>8696198>>8696206>dino
how do you actually do lying besides viewer instead of viewer lying and the girl falling off the end of the bed?
>>8696212You mean lying side-by-side with a male POV? I was trying to do something like that and looks like there aren't any booru images for this in the first place, so I doubt you'll be able to gen that (besides CN from 3dpd).
>>8696212I just do lying on person these days. Cuddling helps too.
>>8696215So I thought. I'm pretty sure I tried gathering a dataset for it before and getting like 3 pics kekWould probably be somewhat forceable if someone made enough of a 3dpd/model dataset to bake over but eh I can't be bothered setting up modelling tools
>>8696220why is her arm fusing into her ribcageand why is her spine so absolutely fucked up
Awwww, dirty n-word feels cuddly. How cute.
>>8696222Resolution. Next question.
1girl, disgust, leg lift, barefoot, looking at viewer
>>8696229Almost have a gen like that
>>8696230I only have gens like that
>>8696232bakaschizo?
>>8696140closest I got tagging the shit out of it and running a lora with the artist style on hires fix
>adding negs actually seems pretty advantageous when using style loras compared to inbuilt artists huh
>>8696236The real redpill is that even this is lora dependent.
>>8696236I said that months ago
>>8696238Is anybody talking? Who's there?
>>8696239hi~
>>8696236Well okay it still softens the lines considerably on a closer lookNot that bad of a thing for some artists though IG.
>>8695667>>8695668>>8695903box please
>>8696243hehe nyo :3
>>8696243is novelai 4.5 :D
>>8696234further refined, I wonder if he just gachas until he hits a really good seed and what artist tags/loras does he use...
>>8696236>decided to do a blind 1:1>18:15kek many such cases
decline?
my fault
I am raping Nobeta, sorry I can't post that here
what caused my chair to recline?
I found the missing piece of the puzzle that was eluding me,taking my gens to the high quality that I wanted. Using a hires fix workflow while pushing the resolution up as much as possible before the gens start breaking down has made the gens so much more insanely good looking its not even close to what it was before.
your lack of generacy of course
>>8696323Yeah that's basically what I do as a 3 year old newfag. Plus gen in hires 1536x1024 from the start.
>>8696170It's iuui + masamune shirow
>>8696325>gen in hires 1536x1024 from the startthat's why her arm is fusing into her ribcage
>>8694130box?
>>8696236>negsNegative prompts?
So here's how you evaluate training results automatically:1. Split your dataset into 'train' and 'test', at least 5 images for test.2. Train on 'train' part.3. For each 'test' image take a prompt and generate at least 3 images.4. Compute FID or SSIM or FID based on eva02 features between 'test' and generated images.5. Pick median metrics.Thoughts?
>>8696439Just img2img of the picture above it, which is not mine. If that's what you wanted, here's the pre-inpaint https://litter.catbox.moe/3r416hasf3i52cwb.png
>>8696475I would rather trust my own eyes, and don't see the point of automating something that only happens once every hour or so. But the theory is sound I supposed.
>>8696487Well the entire point of automating this is to exclude any human factor and various small mistakes that go along with manual x/y plotting. And of course you can also do x/y plots and look at it yourself if you really want.In fact, you can set up a blind test where you play the role of the evaluator by comparing generated images to 'test' images and make a leaderboard. But that's a second step of course.
>>8695154still the best
>>8696475That would just reward frying not generalizing you want
>>8696490Depends, are you looking for the "objectively best" lora in terms of recreating the dataset, or "subjectively best" one in terms of looks? The style reproduction might not work out but it can still produce a cool style, better than the original even.
bots are stupid
I remember textbros once did a pretty big communal blind settings test with public results but lmao @ jmagebros 1) even having someone who can write a a website 2) participating
>>8696506This is why you split the dataset into 'train' and 'test'. You don't train on test images. But yeah, in some cases if the data is too homogeneous then splitting it may not have a huge effect. Suggestions?>>8696507Well, to me this perspective is weird because I can easily distinguish raw gens from human artwork easily, so I don't think "subjectively best" is even applicable here.
>>8696516It'd have to be something simple for one, SDXL still can't into actually replicating promptsBut still, I feel like that'd just be inadvertently become like trying to use loss for quality control, too divorced from life to matter. And how would you even judge what epochs to test? Different datasets converge differently, and it all goes out of wack once you tweak settings.
>>8696519wasn't this supposed to be akame
>check old gens >try to port over old style I love into vpred >can't hit it so try vpred loras same weights on the older model >looks better I swear vpred is a big psyop that I believe in like I can get stuff I like consistently but it always feels like its missing something, or the colors look too soft or something.
>>8696523vpred is a meme because it was made for 3dpd not animeso the ebin chiaroscuro doesn't really make sense in anime which essentially never used it
>>8696520>like trying to use loss for quality control, too divorced from life to matter.How is that so? SSIM and FID are computed on fully generated images in RGB space. SSIM probably isn't particularly good for content evaluation, but it's very cheap and you lose basically nothing by computing it. FID, on the other hand, uses features obtained from Inception V3. Which is not so useful for anime images, but it's possible to swap Inception V3 to using features from wd eva02 or camie taggers (camie was also trained to classify artists).>And how would you even judge what epochs to test?You don't judge, you test each one because all you lose is a little time.
>>8696521Nah
>>8696524>it's either grey sludge of eps or its fried chiaroscuroFuck off.
>>8696524anime noise schedule when?
>>8696524naiv3 is vpred
>>8696525>How is that so?Well, because it won't tell you how it prompts, how it gens different poses, if it has fry in that technically is how the artist does it but doesn't look good etcThe reality is once you have the baked loras with the different settings, the hard part is gone, visual result comparison is the easiest thing to do so trying to automate that out of all things seems a bit silly to me. (t. guy who did the 1:1 test)>You don't judge, you test each one because all you lose is a little time.Okay, but how much are you going to be training? Are you going to be adjusting them on the go depending on the settings or what? I'm just thinking logistics here. You can't very well compare epoch 11/1000 steps or whatever on 1e4 batch 1 and epoch 11/1000 steps on 1e5 batch 2 >>8696528Kek it kinda do be like thatDrawn hentai is unironically pretty grey overall, or well, not as contrast and dynamic range fried as the typical AI is. I would say it's somewhat easier to try to force vpred to look more grey than it is eps to look more colorful for that happy median tho>>8696530NAI3 a meme
sighdo i have to spam this thread too so they stop shilling paypig garbage
speaking of memes i gotta quickly retry that "just bake on illustrious for vpred it's so much better" meme
>>8696531>Drawn hentai is unironically pretty grey overall, or well, not as contrast and dynamic range fried as the typical AI is. I would say it's somewhat easier to try to force vpred to look more grey than it is eps to look more colorful for that happy median thoYou're basing your whole perception of vpred by the poorly executed Noob conversion. A lot of shit went wrong with noob vpred, sadly, and its still salvageable despite that. It doesn't have to be grey eps shit. Fuckton of anime arts are colorful, and the gens dont have to be chiaroscuro crap to utilize proper color range. EPS is nothing more than a technical limitation that is not in present in newer models. And despite your dismissal, what NAI3 did right was colors and looks. It wasn't great with prompt adherence but it looked pretty great, especially considering saas limitations.
>>8696533I may be the originator of that meme. You do lose some accuracy but it cleans up burns and overfitting very nicely, so it's easier for shitters to bake that way. If you're less of a shitter it might be too much going all the way back to illustrious, my best lora was baked on v-pred 0.6.
Also, gotta say, if anyone actually wants to automate settings, it's gonna be much better to do a character as a test.Styles even when well baked are limited by the res and vae and trying to discern whether a slight change in style looks proper divorced of the actual images is nigh impossible.I love testing on Tezuka Rin because she is simple enough to not be broken on most gens but you still get the big indicator of her (lack of) arms and the possible sleeve bleed to compare.>>8696539Well one thing I can see already is that it didn't really converge in the same time the proper vpred bake did.
>try to switch to forge since reforge is old and slow as shit>tile upscaling method on identical settings and models doesn't work anymore, causes distortionalright
>>8696544but at least regional prompter works again, right?
>>8696545what's the point
>>8696548four...five...six. Ah ha! Now I can call him out as a slopper
>>8696535>the poorly executed Noob conversionIt's actually been executed exactly per ZTSNR paper, they also had this problem.
>>8696535More like basing it on static CFG gens despite the v-pred paper itself recommending CFG Rescale. It was the same deal back on EasyFluff.
>try to switch to comfyui>everything just works>more than a year later>everything still just worksalright
thanks for the bitcoins loser
I still don't get where the "lel just use rescale cfg" meme came from. It takes a single gen to know it doesn't really work?
>>8696552But I HAVE to use Euler A Karras, anything else is snake oil!
>>8696559I wrote a much simpler script that limited cfg to 2 or 3 for composition steps back when I used raw noob vpred and that actually worked well. Today I'm just using shitmixes and cfg++ and I don't care.
>>8696566this is literally what cfg limiter does
>>8696551did that paper describe adding cosplayer photos and frying text encoder
>>8696566Isn't that what cfg++ does?
>eps, good>eps not grey slopkekarooooooooo
>>8696574Are we still talking about contrast and color range?
>>8696531>Well, because it won't tell you how it prompts, how it gens different poses,Again, you DON'T train on test images. Test images (usually) have a different pose, unless the artist draws only one pose for some reason. The model doesn't see test images AT ALL, and you compare gens made with the model to test images. If you fry the shit out of the model, you'll get train images leaking on the test images/prompts.>how it promptsThis is literally unquantifiable and exists only in your head.>>8696574>moving goalposts
>>8696579I didn't say anything about test images thoughThat it'll do one thing doesn't mean it'll do another well, and prompting for test images or whatever won't tell you that, unlike manual promptingI know for example that a fempov lora works on base but then just fries when using actual artist loras just because they weren't trained on that
>>8696579>This is literally unquantifiable and exists only in your head.also that's the point lmaoyou're trying to quantify an unquantifiable quality of a thing that isn't made to replicate images exactlyi mean how does an test of "gen test image prompt" and try to compare it using algorithms to a said te(untrained) test image even work? it probably won't even give you the exact same pose or orientation as a test image unless you well, actually overfry it on said imageyour approach doesn't make logical sense to the medium imho
>>8696533Colors kept being fucked up for me. Now vpred bakes work great on eps it feels like too or better
>>8696580>I didn't say anything about test images thoughYou seem to not understand the purpose of test images so there's nothing to discuss.>it'll do another wellThis is exactly why you test on test images which are not included in the train dataset.>>8696582>also that's the point lmaoWhat I meant by that is that you can't reliably tell if there's something wrong with the model either due to variety of factors including but not limited to picking a bad seed or you prompting the model incorrectly. Which results in you having coming to literally schizo conclusions about the model. There's no difference from automated testing in that regard.>it probably won't even give you the exact same pose or orientation as a test image unless you well, actually overfry it on said imageThis is why you are supposed to use FID and not PSNR or SSIM because it measures content, not pixels.
>>8696597If FID can reliably quantify "stylessness" in different posez and resolutions between images then it truly is a godly algorithm. GL, I guess?
Besides, denying any useful data you get from automated evaluation outright because it doesn't cover all cases is just stupid and shouldn't be a reason you don't do it.>>8696599If you want to quantify styles, at the bare minimum you'll have to use a base model for FID which has some knowledge about styles. I imagine eva02 would perform badly in this situation because it was effectively trained to discard any style information. Camie tagger should be better at that https://huggingface.co/Camais03/camie-tagger
>youisn't this supposed to be you doing it
>>8696613just use novelai
>>8696613I don't think ChatGPT can do that yet.
>single particular artist mix fries the shit out with my usual gen settings but it does fine with another sampler/shedule/CFG comboOh hell no... This is the second time something like this happen to me
>>8696641means one of the artists is over or undertrained (probably over)
>>8696523Yeah it's fucking weird man. Loras I train on vpred models work best on illu based models and gives me the improved colors.
only 25 dollars?
>>8696552>recommending CFG Rescaleqrd?
>>8696533>>8696539>>8696541>>8696595Just post your findings and stop memeing. All my illu 01 bakes converged faster than vpred. In fact I only use vpred for characters because I can train it for very long without frying. The vpred style bakes would learn 80% and then fry after 10k steps.
>>8696641dpm++2sa/sgm uniform is all you need
>>8696667no it's not, I had to switch sgm uniform for beta since sgm was giving me bad colours
>>86966673m sde you mean with lots of steps
>>8696670On 102d? I had the same problem with beta giving blurry bad colors which forced me to user euler/beta and I fucking hate that one too.>>8696672>3m sdeHmm...>lots of stepsHow many steps?
>>8696673>On 102d?yeah, dpm++2sa itself is fine but sgm uniform was not giving me the results I wanted
>>8696677What cfg and steps?
>>8696673i use 50 but if you want less go for 36
>>8696665>10k stepsnigger wtf are you doingmy 400 image datasets were 2k steps
>>8696683It's called an exaggeration, anon. The point is that vpred doesn't fry for me, but it also didn't learn as much for whatever reason. Not like this conversation matters since you won't post your settings nor even your results.
>>86966795 of CFG and 30 steps
>exaggerating train steps for no reasonuh yeah i betcha?
>>8696630oh yeah i guess it never said it will do it, it just said some random bullshit and "thoughts??"
>>8696681>>8696685From my quick test (1 picture) 2sa/beta makes things a bit sharper and improved brightness rather than saturation for colors which isn't bad. 3m sde/sgm is subjectively different. Either you like the way 2sa makes pics or doesn't, but 3m sde/beta was also a lot brighter than sgm. I might just do 2sa for normal and 3m for hires fix portions.
>>8696700yeah 3m sde is nice at detailing
>>8696684I'll agree with you vpred is harder for it to take. My settings worked fine for styles from eps to vored but characters I had to like double or triple my rate to finally get it to take
>>8696667I need to just fuck around with latest reforge I guess so I can properly decide if your meme'ing or not missing all these samplers
>>8696663https://arxiv.org/pdf/2305.08891#subsection.3.4>We find that as the terminal SNR approaches zero,classifier-free guidance becomes very sensitive and cancause images to be overexposed.After which they propose CFG Rescale as a solution. These days we also have CFG++, dynamic tresholding, cfg limiter, etc and even manual cfg scheduling. Point is to have very low CFG in the first few steps, then ramp it up to a normal value.
I can't believe the real answer was to use eps 0.5 all along
>>8696704>with latest reforgeI heard someone say last time that 2sa was removed or something? 2sa has been there since the beginning so it's not "new" or a "meme."
>>8696706i unironically would if someone shitbaked it properly
actually i should try replicating 102d myselfi now know how
>mixes>epslmao
>>8696700Hmm I don't know. They're both just as fuzzy but slightly brighter colors? I can just use cd tuner on the hires fix pass to improve colors. This will be a lot of work to test.
how do you get the girl's head to lay down on a surface? in this case a bed
Try the comfy versions of samplers if you haven't before, they're way better
>>8696735head on pillow
>>8696737then that's the pillow not the bed. I want her to lay completely flat
>>8696738head on bed
>>8696736Hey friend it seems that you were so excited to share your findings that you forgot to attach the X/Y plot proving your statement about image generation! Don't worry, it happens to the best of us and we are eagerly waiting for you to share it so everyone can see it for themselves ;)
>>8696741What are you poor or something? Just do it yourself.
>>8696740Wowie anon let's not get ahead of ourselves here.That kind of thing... It would only work on NAI.
>>8696743apparently, she's sleeping with her head straight up in the air, this shit is so fucking busted
>>8696742Wow no need to be rude, but that's just how this kind of discussion works you know? You provide some evidence to corroborate your claim and then people are able to judge for themselves.If you could just make unsubstantiated claims and then ask people to do the experiment themselves, it would be chaos! Also, it would open up the doors for trolls to spread purposefully malicious information, and that's not something we want in our community!
>>8696748I am not discussing anything doe? Is a fact.
>>8696749But anon you didn't provide 「facts」just 「opinions」, 「facts」require evidence to substantiate them, otherwise they are 「conjectures」 or「opinions」and you cannot plce the burden of proof on others. Hope that helps!
>>8696752>and you cannot plce the burden of proof on othersProof?
>>8696743>>8696745head down
>>8696745>her head straight up in the airPic?
>>8696757Kek.
>>8696743>That kind of thing... It would only work on NAI.It does!! People were shitting on T5 a lot but honestly it feels like the thing that carries 4.5 despite its very meh aesthetic
>>8696745I can get it pretty reliably, but only with negpip (pillow:-1), else it will always show up under her headjust on stomach, on bed, head down
>>86967451.5kino...
>>8696772What about from side, on back, profile?
>hey wtf base noob with rescale really does look way better>realized the model didn't update and i was still on 102djej
>>8696772I negpipped pillow and the whole bed disappeared, now 1girl is sleeping on wooden floor.
>>8696778Some of you still deny it for some reason but that shitmix is legit good enough for most part, perfect for genning
>>8696772same prompt and seed, just added your stuff https://litter.catbox.moe/iy0yxp0ychsl9v3u.pngneg head_rest too if you want her arms out of the way, or better place them somewhere else intentionally
>>8696778yeah definitely doesn't make base usable out of the boxgood for 102d though, definitely helps fix that vpred fry-in even when using low cfg, i'm keeping it on>>8696780it will always mog others just by the fact of being loras instead of cattower merged with easyfluff and wai 10 times over like the others
>>8696778So, does base not look better with it?
>>8696788>>8696784i guess a bit regarding the contrast but it still doesn't make it more detailed or anythingthat's with a style lora
>>8696776>>8696779What do you mean? Like this?>>8696778lollmao
>1.02d page only has the latest shittier version recipe>civit not on wayback machine>no real merge metadatan-not like this..,
>>8696784catbox??
>>86967941girl, sitting
>>8696795I wanna see your rescale settings, obviously.
>>8696772just negpip it for the first 30%, same effect but doesn't bork your whole gen most of the time[(pillow:-1.0)::0.3]
>>8696796the default out of the box
>>8696793lollmao!:skull:
>>8696784Not a great comparison, they all look way too similar. A strong artist maybe, or heavy quality tags, or a lora?Here's mine. Base, rescale, cfg++ and 102d custom
>>8696789That's what people said about loras stabilizing noob, you don't need anything else once you're using one even at a low strength.
>>8696805isn't it the same with yoursrescale on base only added hand melties everywhere
>>8696805>102d custom is just an overpowering garbage style from all of the merged lorasI will never understand shitmix users
loraless it doesn't really change that much either on baseyeah sure 2% better but it's not 102d-usable tier to me
>>8696814base looks like looking at a miniature at 300% zoom with how shitty and unfocused the lines are
>the reason people use shitmixes is because they can't bake their own shit>and don't know how to find loras that they actually like>so instead they just rely on whatever shitmix has the strongest built in (AI-generated:1.21) output they can findeverything was a mistakemaybe there hasn't been any image gen advancement because you heathens don't fucking deserve it
>>8696810why is she wide eyed?
>>8696820Good. Only you deserve to make good pics and those pics are for your eyes only. You've ascended.
>>8696040Seeing her anime-fied looks weird to me is all
>shitmix baitwtf have i taken a time machine to may?
>>8696814>>8696817In the end all that matters is how nice it looks not how faithful it is to the style. It's like that other anon talking about eyeballing epochs. Your goal is to pick out the one you like the best and sometimes it isn't the one that's 100% faithful.
>>8696808at least the XY shows it doesn't do much compared to other options
>>8696825Can you like not reply to me with these nonsensical non-sequiturs? Thanks!
>>8696828Huh?
all that matters is that it's a placebo piss-water that only changes a few small details while retaining the AI generated quality we've all known for so long and grown to deeply respect and love
>>8696820My style mixes looks almost 1:1 on base noob against 102d, the thing with base noob is that I need to roll way more than I should for a good base gen
>>8696821it big
>>8696834WHY IS MR T. WHITE?!
>>8696828Ohhh... you really are the worst shitposter here. Even my shitposting is higher IQ than this kek.
>>8696835ouch
>>8696834zangief's knuckles are fucked
>>8696837that's what communism does to a mf>>8696841Hmm, true, I didn't put too much care into it at that time
>>8696834WHY DID HARRY DUBOIS GET A MOHAWK?!
god damnwhatever anon mentioned rescale i thank you
>>8696847cheers
>>8696847jesus talk about a conclusive blind test
>>8696849what are you using
>>8696850https://github.com/rainlizard/ImageBatchCompare
>>8696854wtf that's great
>>8696856quitei'm pretty sure an anon made it too, i think he shilled it in hdg or something and that's where i picked it up
>>8696854that's actually really cool but i also wanted to know about your rescale settings
>>8696860>>8696798
>>8696864What happened to Serena’s clothes?
>>8696868Lost them
okay one bad thing about rescale is that it does quite exaggerate the background greyness i get with some character loras
>>8696878not just with loras, lol
>>8696880it ain't that bad without chara loras at leastdefinitely goes full on grey on white backgrounds with characters, i wonder why
>>8696882nah dude i'm runing x/ys right now and rescale is consistently grey
>>8696884me too and it's not that bad lol
oh it's when it's a chara lora with a style lora that it shits itselfeither alone are okayweird stuff
>>8696887when i turned the cfg up with rescale on i got distortion that looked like when you got too many loras applied so i presume it operates in a similar way somehow
>>8696890From what I can see the difference with rescale using one lora is literally nothing backgrounds wise, but when I put in two+ of any loras it just utterly breaks into grey meltieHow weird, I bet that's some kind of a bug.
What if I was already using the skip early cfg setting? Why bother with rescale?
>>8696475Nigga just now thought of validation splitting. It's been part of ez-scripts for a year, lil' bro
>>8696909yeah i already use guidance limiter (same thing pretty much) and it's nice
>>8696490machines aren't qualified to judge artistic value
>>8696909>skip early cfgoh man that's fuckin weirdwhat value do you use?
>>86969150.2
>>8696916real strongit's the ignore negative prompt thing in settings, right?>>8696912gotta check it outrescale is definitely bugged
>>8696918>negative promptlollmao
>>8696919>Ignore negative prompt during early sampling (disables CFG on a proportion of steps at the beginning of generation; 0=skip none; 1=skip all; can both improve sample diversity/quality and speed up sampling; XYZ plot: Skip Early CFG)
>>8696918stop using negative prompts
it doesn't matter if you have a negative prompt in you fucking retards
>>8696919go back to trying to shill 6.8gb lora refiner shit
>>8696926not for guidance limiter/etc no but it still doesnt look good
anyone have recs for artists with simple styles? attempting to shit out a synthetic dataset for a character
>gotta get in my le opinion (shared by 99% of the thread) even if it isn't relevant to the discussion xdddd
>>8696921Yeah>>8696927I was memeing about you using no negs/neg pip instead. Why are you so mad?
>>8696931this isn't twitch dog
>>8696925what the fuck is it from anyway? reforge has shit fucking extension naming
>>8696935it's one of the built-in ones and is combined with "APG's now your CFG" (useless extension)
>>8696930ixy
>>8696936so it is, TYman both of them along with the skip cfg look kinda ass compared to rescale with how much they change the gen compared to rescale but rescale is bustedmaybe i can get gpt to try to fix it kekor just won't use two loras at the same time, i rarely do anyway
>>8696940don't use both guidance limiter and skip early cfg, they are almost the same thing and both together will fry the shit out of your gen
>>8696941well yes of coursei mean either look ass
>>8695607>>8695607>>8695607
>manually fine tuning your sampler/schedule type/CFG per style mix >> layers of snake oil
>>8696932>Why are you so mad?He got buckbroken by a funny poster over a year ago https://rentry.org/anypony-science
can you likeinclude references so i can know what you're making fun of and if i should be mad
>>8696946This but ironically. Better to spray and pray.
>use eps 0.5>use euler>cfg fixed at 7I have reached enlightenment
Ohhhh I member refine guy, he stole some of my NAI gens for his experiments (which wasn't that bad tbqh)
>>8696847Where is rescale/ give settings
>>8696955>eulerI will come home soon, brother. But not yet. I'm certain these samplers have value.
>>8696958SHUT UP SHUT UP SHUT UPPPPP
Has anyone switched reforge branches to see wtf is up?
>Nigga just now thought of validation splitting. It's been part of ez-scripts for a year, lil' brouh-oh, both are false!>>8696914Neither you are
Remember when everyone was using a model where quarter of the dataset was my little pony art and another half sourced from deviantart, and it had Reimu in promotional posters because it was 1 of the 2 touhous it could barely generate?
>>8696965You stopped?
>>8696962I think I am on the latest reforge main branch before pancho decided to quit (again)
>>8696963neither are you*
>>8696965what was the other 2hubecause I don't think it was able to do the terrorist and only really remember it doing the pit slut
>>8696963kek this place will never not be about shitflinging
>>8696965Grim times honestly, I am glad we all moved from there
>>8696970Flandre (with the wings and shit).
>>8696972Deadnoodles worked just fine on there too, cuckie.
>>8696974huh. that's a weird choice considering how hard the dumbass tried to prune lolishit
for me it was the model baked on 512x768 garbage that could not do any character or any style and was fried with literal 3dpd pedophilia to make it not melt we loved that for a year+
>>8696975It was the lora the former oreocookieschizo did for me and it was on top of reweik, real anime concepts and styles didn't work well
>>8696971you asked for it, didn't you
>>8696979No idea what you're talking about and I even remember when you asked the thread to bake the lora for you. Yes reweik was peak and things weren't as bad as you remember them. You're right about concepts though, I remember anon struggling with that upside down deepthroat and we couldn't get pov lap sex to work when it worked on nai.
>>8696965Just finally deleted base and Reweik last week, realized I haven't used them in months. Still keeping PonyRealism.
Oh hey nice Forge Classic finally attained essentially output equivalence with reForgeNow I just need some of my samplers
>>8696997How fast is model switching on there?>meme samplerseuler is all you need
>>8697000i dunnyo but loading the actual program takes like 3s not 30
>>8696986What I was trying to say is that just a very few styles tags worked out of the box, same with character tags, while it was """acceptable""" I was praying everyday for a real anime-oriented model to come out and take over Pony was honestly fine for /h/ gens but for me was lacking many things for 1girl standing kino
>>8697002Hmm... guess it's worth a try.
genning's a bit slower on classic thoughi assume because it tries to force sdp attention
>>8697014>slow genningDropped :/
>>8697020who is this non-loli
>>8697022i just typed stuff into the prompt
>>8697024so based
>>8697024>>8697020Box??? What's the tag for the clothing?
>>8697029I would guess cleavage cutout
>>8697031No like the colors and stuff.
>>8697029https://files.catbox.moe/2ny9iu.png
>>8697037>holographic clothesThank you.>lolicomkek
>>8697039it's a nice artist
Hey, AMD + Linux user here. Now that pancho stopped updating ReForge, I don't have anywhere else to go, since the alternative he's giving me is Forge classic from haoming02, but that webui doesn't supports Linux. And I've tried illyasviel actual Forge but it's giving me serious perfomance problems. Is there any other alternative that isn't CumragUI that is more like ReForge?
>>8697044tell me literally a single thing you expected to get out of any future reforge update that you don't have now or prevents you from using it
>>8696486very cool seeing the pic I made transform into this,which lora/artist combo to get this style of genitalia?
>>8697044you can use swarmui if you really don't like comfy's ui https://github.com/mcmonkeyprojects/SwarmUI
>>8697047just https://civitai.com/models/1521385/qwert6969-artist-style on 102d customurethra is inpainted manually
>>8697052tips on indrawing small detail? whenever I try to indraw something like that either nothing happens or it gets completely garbled up
>indrawingholy newcutie
>>8697056Pickup a pencil, not a brush.
>>8697054NTA but first and foremost, check your inpaint settings, you should be using the same sampler and CFG value as your base or upscaled gen
>>8697054you're probably not using only maskedgotta use only masked for details
>switch to forge classic>change defaults>doesn't save defaults>look at img2img>go blind from noisy checkerboard>gen image>looks bad and takes longerWhy the fuck did I even waste time downloading this shit?
do forgekeks really
>abandoning reforge to use the same techyou guys have problems
>>8697066Exactly
>>8697065>>8697066>>8697067To be fair I was only doing so so I could prompt for backgrounds with flux and gen pics of my waifu without opening two separate cmds. If reforge would support flux I'd never look back.
comfy still just works btw and supports everything
>>8697072>>8696889
>>8697074>using bloated slop workflowsdont download your shit from civitai. you dont need extensions when you're starting
would it be a good idea to go through a ksampler 3 or maybe even 4 times everything with the same settings but tuning up the cfg each time?
>>8697076
>>8697079nyo>>8697080el basado
>>8697082>nyotoo much cooking? I guess a double pass is already enough to detail everything>>8697080been learning this stuff for not even 2 weeks and I already quickly realized the more bloated my workflow got the worse my results
>>8697045More snakeoils? Support for future subtypes of LoRAs? Support for future extensions? You tell me>>8697051How is it? It looks way different compared to Forge
black souls lora onegaishimasu
>>8697080this but unironically
>>8697087>How is it? It looks way different compared to Forgeidk, i've never used it, comfy mostly just werks
>>8693972>FOR DOING GENS:>- 102d custom| https://civitai.com/models/1201815?modelVersionId=1491533>- 291h | https://civitai.com/models/1301670/291h>- r3mix | https://civitai.com/models/1347947/r3mix-illu01nai-v-pred>BROOOOO JUST USE THIS BASE MODEL I MERGED WITH MY SHITTY LORA ITS SOOO GOOOD BROOOOOOOOO
what caused the base model meltie?
yes
>>8697101>BWOOOOOftfy
>>8697101custom does suck, original is nice
>>8697101uh chuddy you know you can just bake without it? it's some random botty retard that did that and didn't even bother changing the template link
>>8697054Open up inpaint sketch or an image editor and draw what you want to see, in rough blobs of the correct color. Then run a low denoise inpaint over it. This is the most reliable way, if you don't mind the extra effort.
>>8697109say if I wanted to do something tiny like telling the AI to draw tons of freckles over parts of the body without modyfying the body itself how do you think I should go about it?I tried painting over the area but it changed too much and painting a ton of little dots not only took a decade to load but basically did nothing
>>8697113mask the face low denoise prompt for freckles
>>8697108>didn't even bother changing the template linki forgor
>>8697113That's a weird use case, I'm not sure. You could mask only the flat areas and not the outer linework, that might keep the body from changing. They shouldn't need much denoise anyway if they're small dots.I might give it a try next thread.
>>8697114I might try again but I did that before and it resulted in either too much change or no freckles being painted at all depending on denoise strength>>8697123do post results, I'm thinking in this specific use case it might be a better idea to learn how to do it in photoshop if anything.
>>8697102Residual resentment.>>8697109Thank you for your explanation.>>8697113>frecklesBro no fucking way. You shouldn't even need much work for this. Body freckles is the tag your looking for and raising the weight will get you 90% of the way there. At worst yeah just draw them yourself, mask the spot, inpaint only masked, then prompt 1girl, body freckles, and ur done.
>>8697126In hindsight you should've provided the style mix, will probably influence how difficult it is. Or even the exact image in case there are other effects like weird lighting or whatever.Anyway I took a random ginger 1girl(1), masked her like this(2) and added body_freckles tag, 0.4 denoise(3). More is overkill, they're just small dots. Face seemed insufficient so I rerolled it with higher tag emphasis(4). No freckles on the belly makes sense cause it gets the least sunlight normally, but hey I added some manually(5) and rolled a low 0.2 denoise over it(6).Only issue is, cause they weren't all inpainted in one go they ended up different sizes.
>>8697136catbox?
>>8697147facial freckles worked fine for this task but it refused to draw lots of freckles through the body, it kept to patches on the shoulders and thighs for whatever reason but I'll try it again later in a different picture now that I also significantly upgraded my entire gen workflow.>>8697149not posting the image cause futa but I used only zero_a as an artist tag,its a friends OC with tan skin and lots of freckles all over the body. I'll try doing something like that to see what happens when I try genning her again.
>>8697158Why not draw them manually (it's just dots after all), and then img2img at like 0.35 denoise with body freckles tag? Skip the whole masking and indrawing all together.
>freckles>futa>a friends OCdoes it get gayer than that?
freckles are nyot gay you heathen
go back
she's a local
>>8697163Yes, it does
bara gay furry
>>8697162I suppose I could try drawing rough dots on photoshop or something and then try the indrawing method>>8697163I mean I also genned my other friends femboy oc taking dick so I guess? GAY for REAL would be something like bara hairy dudes
>>8697188>my other friends femboy oc taking dick>>>/d/ is more your speed
Having the first ksampler cfg scale super low can give some interesting results>>8697191none of the other threads have as good discussion on the subject of genning, they're pure goonery so I default to here
>>8697197I didn't realize that was still true after the decline.
>>8697200I'm new around these threads so I wouldnt know how it was before the supposed decline but the other threads are essentially people just posting their mediocre slop to fap to and no discussion at all
>>8697200there's techniques being used in /trash/ that puts us to shameunfortunately it's used to generate the most degenerate shit known to man
>>8697155>https://files.catbox.moe/r96zw0.png
>>8697208>there's techniques being used in /trash/ that puts us to shamelink them, you said the same about /d/ gens the other day
>>8697217>VAE: lpips_avgpool_e4.safetensorswe got a scientist over here
>>8697208/slop/ has some real effort genners but they're buried in the bellcurve slop of the same 5 dudes screaming for everyone who is offering to gen their shit waifu which somehow is always asuka, rukia, or one other I can't think of off the top of my head
>>8697219>you said the same about /d/ gens the other daymedsanyway last time i checked in there some people were trialing this: https://civitai.com/models/1233692/style-ipadapter-for-noobai-xlbasically vibe transfer at home. works really well on some styles, never seen it mentioned outside of there
>>8697223>https://civitai.com/models/1233692/style-ipadapter-for-noobai-xlthe insane tech is....controlnet...?, not sure about the others but I have tried this one before, way too slow on my end with very poor results
>>8697229>custom ip adapter>controlnetbro?
>>8697233yes?, feel free to explain and provide examples of the differences between them
>>8697223>the super secret tech is....>vibe transferthere are people who post in this thread that will literally make a style lora 50 times while making minor dataset tweaks you dumbass
>>8697235it's not technically a controlnet but it effectively does the same thing
>>8697236>there are people in this thread that waste their time attempting to reinvent the wheel to no availthis is common knowledge
>>8697219That was me and it was also controlnet. The point is since they're doing crazy /d/ shit they have to get creative to achieve complex poses or insertions while "vanilla" /h/ is just benis in vagoo which you people hate to see after a while.
>>8697217>dpm++sa/betaHmm yeah I'll stick to sgm uniform for now but I see why you'd use beta. It will depend on the style/ for me.
>>8697246beta is underrated so is euler a DY cfg++
>>8697221Sakura.
>>8697223I've actually been messing around with that lately, posted some rough experiments on various boards. >>>/e/2992081It's difficult to find the right settings so that it transfers more of composition, style, background or character details, usually it just takes everything. And it's nowhere near vibe transfer for character details, can help and undercooked lora that's about it.
>>8697252interesting, in my experience it usually only attempts style transfer, less so for characters/composition. it is pretty hit or miss as to what styles it recognizes, needs to be a pretty high weight too.
>>8697257Well it's weaker than your prompt, so if you already have a pose or character in there it'll only touch features you didn't specifically mention. Here's an extreme example.Might also be that Comfy uses different default settings for ipadapter. There's some stuff here, a weight type, a scaling type, start/stop steps, area mask and strength.
gyat damn i need to do more nonstandard res gensdoin some 5:2 landscapes and it's kino
>lil bwo is finally catching up
>>8697281>nonstandard res gensWhy were you laughing at me when I told you that, shitposter?
>>8697283AHAHAHA IT'S THE SMELLy PANTS LOSER GO BEFORE I SHOVE YOU IN A LOCKER HAHAA
>>8697284>SMELLyhow did this happen
>>8697285the material conditions
Anyone got the name or link for that tool for datasets that can apply backgrounds to png's and did other shit?
Trying to test what I can do with WAN 2.1 now that someone has distilled it and made it run on 8gigs of VRAM in a reasonable time frame. Can this model actually understand sex or do I just have to /e/-prompt girls shaking their asses? I'm using the i2i version so I at least have a base image to work with.
>>86972951. use wan 2.2, it's less censored than 2.12. you'll need a nsfw lora, thankfully that exists for 2.2 and 2.13. if you have questions ask in >>>/g/ldg and not here
>>8697293Dataset editor. Should be a link in the last thread.
>>8697298>1. use wan 2.2, it's less censored than 2.1I tried using the lightning version of 2.2 and for whatever reason ComfyUI kept crashing when it came to one of the nodes. Maybe my workflow was just fucked but I got it from the Huggingface page. Is it that much of a difference?>2. you'll need a nsfw lora, thankfully that exists for 2.2 and 2.1I'm assuming you can get these from Civitai?I'll direct further questions to the /g/ threads, thanks.
>>8697306>kept crashing when it came to one of the nodesare you by chance using the FP8 versions of the models? id recommend using the GGUF ones, specifically Q8.
>>8697310I don't think so. The models I downloaded were named>wan2.2_i2v_high_noise_14B_fp16.safetensorsSo I imagine they're the full fp16 precision. I might try following the Wan2.2 guide from /g/, it also mentioned those versions.
>>8697316yea id still rec using q8 unless you have a RTX PRO 6000 or something
>>8697318Oh, that would explain why it wasn't working, yeah these models are way too big for my system to handle, even with virtual VRAM. I'll grab the quantized stuff and set up a proper workflow. Thanks again.
>>8697303Feel like an idiot I added background color to the transparents it shows it in the program but none of the transparents have it in windows explorer, so I don't know how you check it.
quite satisfied with my results
>futa cock
>>8697340please stop disrespecting 1boy like that
is the futa cock in the room with us
hmm I haven't done futa in a while
why is /h/ so obsessed with a /d/ fetish
>>8697331Did you try saving? I haven't tried that feature yet so idk how it works.
>>8697339What is the base res for this?
>>8697353Yeah I knew I was retarded I had to select all since it's a literal >apply backgrounds to selected imagesTime to re-bake this see if it turns out better. Only thing I'm trying to figure out now is I re-tagged the dataset and added the activation word to it but I can't figure out how to make that word be at the top of the list for images.
>>8697355I'm planning on using this in about a week so I still haven't learned all the features but for this one you type the position first (pick 1 instead of -1) and then type your tag in the global tags on the right then hit enter.
>>8697366Bless, sure it's in the documentation I just missed it I figured default -1 would place it at the very front for some reason.
>>8697355>Time to re-bake this see if it turns out better.if it's just adding random rgb values and you're using an auto tagger to caption it then it's gonna turn out worse probably. i'm not really sure why. i tried using booru dataset tag manager's random color fill functionality and it wound up turning black clothes blue or red. i wound up just making a script that detects if an image has transparency and then picks a color from a table and automatically adds [color] background to the caption file. that's worked out fine so far.
>>86973541440x2560 picture scaled to 0.75 megapixels passing through the first ksampler then 2x latent upscale through another ksampler img2img workflow
>>8697401upscaled 2x again at the very end I forgot to mention
>chadmix faggot is still aroundgrim.............
lil bro is still fighting ghosts:skull: :skull: :skull:
he just put some garbage look >>8697106grimmsly..............
>>8697298at least in my experience wan 2.2 is fucked for 8gb vramlets compared to 2.1 because of the dual model loadingunless something better has come out since i tried it 2.2 fp8 doesn't work, only 2.2 gguf and that looks way worse compared to 2.1 fp8
big decline
>prompt flat color>hair is still shinythis vexes me
i too am in this thread
>>8697540you are black
>>8697538>prompt shiny skin>face is matte
>>8697538>prompt best quality>it still looks like aislop
>>8697455i myself am an 8gb vramlet, with 32gb ram. i use Q8 (which is the only worthwhile gguf quant) and it works with both models. it is slow, 5~ mins for the entire workflow including interpolation. if you were using a lower quant like Q5 id understand but Q8 is the same if not better than FP8
I want to throw some realism into my gens, but without using just the tag, its too random, do you guys have any suggestions for artists that are realistic, but with anime proportions? unlike sakimichan and stuffit's hard to search for artists lol
>>8697626>it's hard to search for artists lolIt isn't, danbooru gives you a way to search for related tags. You are just lazy and retarded.
zased
>>8697626i agree with other anonthe tools are there
>>8697626we're in an age where you can vibecode a scraper that does exactly what you wantget fishing bucko
>dip for 2 days>hgg was actually active>scroll through>schizos>vaseline drenched melted gensWell, guess it saved me having to also check /hdg/
>>8697701>vaseline drenched melted gensnothing has changed for over a year, why is this notable
>>8697701>2 day breakfucker is an addict lmao
imagine /hgg/ being your main thread
>>8697626s16xue
Sorry bros, I finally ran out of drugs, and withdrawals are terrible. You're gonna have to do without my gens or advice while I am recovering.
>>8697626>realistic but with anime proportionsNow you made me want to bake like miku doll style
>>8697626nixeu is my go-to
>>8697626guweiz
do you guys use anything but grabber to scrape boorus?
>>8697701>vaseline drenched melted genswhere
>>8697745No, it's good enough
>>8697715i actually baked a bjd lora before, it was undercooked thoughhere's the dataset for just the vinyl bjd part if you wanthttps://mega.nz/file/4qJzxCLK#AmOycFoQzNGvlW4ROUf_M03MEY8T6KE6Z0A0bk-oyx0>>8697745i've used gallery-dl for pixiv scraping before but for boorus i just use grabber
>>8697745my scriptgrabber got all fucked up recently for me
>>8697745i tested and it actually takes less time for me to manually page through because a huge percentage is actual garbage
>>8697753Thanks I will give this a go.
>>8697755But you can do the manual selection in grabber.
i don't know why i baked this but i did https://files.catbox.moe/omvh8c.safetensors
>>8697758migu lora?
Anon has convinced me to save up for a promax 6000.
>>8697762Is that like a Fallout vacuum brand?
>>8697757not that i sawbut the ux is fucking dogshit in grabber so i may have missed it
>>8697771>ux is fucking dogshit in grabber so i may have missed itYeah I'll give you that. But you just click on a pic, or control + click I don't remember, then click on get selected and it adds it to the bottom on the selection page not the top. Note that it only works per page so if you leave the page you lose all your highlighted pics.
>>8697755That's a bit insane with how much manual clicking that'd need (remembering my oldest lora bakes kek) but you're right about the garbage and that's why I love baking only on danbooru when possible.
>>8697773but isn't that the same number of clicks as just using the webpage? i fail to see the point
>>8697819No, no, she's supposed to rub her labia with the sword not insert it into her pussy but everything is a dildo if you are brave enough I guess
>>8697745hydrus
>>8697758fake a fart lora sars pls
>>8697835She already past that phase of her training.
>>8697904Nice, very based
>>8697819consensual stomach bulge is my fetish
kohya_ss can't configure accelerate anymore. One trainer is too slow (does random number of steps). I guess I gotta paypig for training..
testing https://civitai.com/models/1790792?modelVersionId=2122326not great at porn but i'll probably use it for sfw gens pretty often
>>8697987it's good with natural language without requiring it which is nicecompare my attempt at this image https://danbooru.donmai.us/posts/9529872one being just the tags and the other tags + sentenceshttps://files.catbox.moe/oxjx1s.pnghttps://files.catbox.moe/yu57w5.pnghttps://files.catbox.moe/2i8qv1.png
>monochrome>greyscale:1.2
>>8697981>using some retarded gui>random steps???jajwai
>>8698000So whats the best lora trainer then?
>>8698003if you can't configure accelerate then i can't help you because everything's sd_scripts
>>8697987I'm going to download comfy just to delete in after one gen again aren't I?
>>8698005A year ago I was about to, now its this: https://files.catbox.moe/hbzcr1.pngPress 5 and stuck. Keep smashing 5 out of frustration..
>>8698007if you develop masochism then you can start using comfy 100% of the time>>8698008try the iriginal sd-scripts insteadhttps://github.com/kohya-ss/sd-scripts
>>8697981why do you need accelerate in the first place? are you doing multigpu training?
>>8698011Its taking forever to cache latents and telling me I can't use xformers.
>>8697995Okay that is admittedly a tiny bit better than I expected. 50s per gen better? No lmaowait no scrap that i just got the second output and holy overbaked on waislop outputs
>>8697987>How is his support for women with plump European and American bodies?oh no no no no HAHAHAHAHAAAAHAHHAHAHAHAHAHAHAHAHAHAHHAHAHAHAHA
>>8698015???you can adjust vae batch size and/or split the dataset if all you need is to cache some latents https://github.com/kohya-ss/sd-scripts/blob/main/tools/cache_latents.py
>>8697987there are still people who use this lying shitmixer's models? he doesn't train shit but pretends he doesit was hilarious when he rebased to rouwei and minty got it to gen his sikrit watermark
>>8698016funny it doesn't feel that much slower then sdxl for me but that's probably because im on 8gb so there's no where but down for meaccording to the description it's just trained on danbooru and e621 but i wouldn't be surprised if they were lying and there's a big dump of random civitai outputs in there
>>8698015xformers shit is more because the requirements are all fucked out of whack for modern setups, at least on linux
>>8698031I don't remember if it was Neta or not but one of those base models was definitely FRIED on WAIslop for anime, might be bleed
>>8698029>mintyliterally who? a friend of yours?
>she doesn't know minthyholy newcutie
>>8698041literally the person who made the model this nigger ripped off and attempted to pass as his own train:https://civitai.com/models/1308285/illumiyume-xl-illustriousread the comment section for the lulw
I'm trying noobai-xl for the first time and a lot of my gens look like this. Am I not prompting right or is something wrong with my forge?https://files.catbox.moe/ugm2gd.png
>>8698029it's mildly better then the base netalumina so even if it's just a lora slapped over it's good enough for me>>8698050your computer is about to explode you should get down
>>8698050You need cfg limiter for vpred. Or cfg++ samplers. Or use one of the shitmixes (102d, 291h).
>>8698050Aside from the questionable prompt, try to use the CFG++ samplers first and change the value of the CFG to 1.5But honestly if you are new into this you should be using another model
>>8698050>>8696705
>>8698060>>8698058>>8698057Okay I will look this stuff up, thanks
>>8697987can it do miku doing a sitting split on a purple cube?>>8698047i wouldn't be surprised if they both rip each other lol, just look at rouwei model page
>>8698047I think the funnier thing in this is that people are genuinely that pressed over citivshimixes.
>>8698065Just cause it's using the wrong category, when you search for finetunes you don't want to see merges.
>>8698064>if they both rip each other lolNo.I really dislike the rouwei models, but they're legitimately trained by that autist and behave very differently from your average shitmix. For the worse, if you're not into /aco/ or cold color palettes. IMHO they're failbakes and it's a sign of Duongvechinkoid's lack of intelligence that he chose to rebase to that model instead of sticking to noob which was what he previously used as base for his "I actually trained this" bullshit mixes.You see, RouWei is trained from Illu 0.1. It has zero noob in it, which is another reason for its inherent inferiority.
>>8698069Oh right he is that retard who "finetuned" Noob and it's basically a 1 step tune because it has less difference than a low strength lora.
>>8698065>people are genuinely that pressed over citivshimixesit's rare to become like this and it was entirely caused by the author of the shitmix writing shit like this:>The model was trained on the danbooru2024, danbooru_newest-all datasets, as well as a custom dataset (which I collected and labeled using natural language with GPT-4.5, and later manually verified by me).>This is the first base model I've created, so any feedback is welcome>first base model I created>base modelGrand claims cause grand butthurt when they're bullshit and you're nothing but a master of ctrl+c ctrl+v, it's the way of life.
>install sd-scripts>uses old pytorch from 2 years ago>doesn't have gui.batvery epic
>>8697987Could you please run the same prompt+seed comparison between original neta-lumina-v1.0-all-in-one.safetensors and this finetune on a couple of prompts?
>.batoh nyo nyo nyo nyo
>linuxnot even once #batforlyfe
>>8698077why would sd-scripts have a gui it's a cli program>>8698078i already deleted the original so give me one sec
batty boy
>trannux
>>8698081I'll switch to troonix when it supports VST plugins (thankfully this means I'll never have to switch).
>>8698083>why would sd-scripts have a gui it's a cli programso? CLI programs can have guiI personally don't see a need for them but for eghttps://www.opbyte.it/grsync/screenshot.htmla GUI for rsyncit's a thingsome people like clicking checkboxes and having file dialog open for paths more than reading the Fucking Manual and typing --argflag
>VoSTok pluginsdamn russkies
>>8698090yeah but you shouldn't assume it would have one
>>8698064i got distracted plapping chatbots sorry about thatit can put her on a purple cube but it can't make her do a normal splitalso i was wondering why it wouldn't give her twintails and it was because i typed rwintails im fucking retarded
>>8698032Why doesn't it use something newer like sage attention?
>>8698100>quantized optimizationishiggydiggy
>>8698069>>8698076desu you're not doing a great job at preventing these idiots from becoming popular. i somewhat agree with the other anon, there's no such thing as bad publicity. same thing with resonance, pony v6, etc
is this lil nigga a twitterite
>>8698096damn this looks bad>>8698032anon you're doing something wrong, it's piss easy to install xformers nowadays, especially on linuxhttps://github.com/facebookresearch/xformers#installing-xformers
>>8698110you can make it look alright if you use an artist style but by default it doesn't look great
>>8698110anon i said requirements not xformersthe sd_scripts requirements.txt are old as shit, install old ass versions that don't work for modern pytorch
>>8698113i mean it doesn't even look like a cube, it's a fucking rectangular prism>>8698114you realize you can just install the new version on top of whatever shit it came with?
What's the point?
yeah anon is right i just don't wanna talk with a retard that just wants to have the upper hand about irrelevant shit
>>8698114also this is why people use uv or pixi, get over venvfags
>>8698118It was slightly cube-ier in a previous gen but good point.>>8698120My dick.I'm a dog.
>>8698122>begs for advice>cries when people actually give him advice
>everyone in a general is one personHELLOOOO SCHIZOO
>bake 300 styles>use 3 on the regularIt doth be like thats.
>>8698136Yes? I thought we have already came into this conclusion, all of you are (Me)
>>8698142upload all of them right now
>>8698144NYA i didn't even get yous last time
>baking loras for (You)snew low
how do ppl fail at the most basic logic comprehension here
what is schizobro arguing about now? I just got in here.
Lil niggeti thinks uploading loras is baking or what?
>>8698078i got distracted sorryhttps://files.catbox.moe/kh8cbx.zip
>>8698167No problem no problem, thanks for the work. Looks more detailed across the board. Original is more like vague blobs, finetune produces more complete images.
>>8698169yeah i think the cheap looking default style is a fine tradeoff for more finished looking images overall
k I need an adult to give me a .toml for characters with illustrious this shit doesn't work at all most of the time.
>>8698209https://files.catbox.moe/4gkbpf.jsonSource is /hgg/. Use the newer illustrious1.1 to train also works. Also working with Batch size 3 if you got VRAM to spare.
any tips on getting rid of this stomach crease? inpainting is either doing nothing or reinforcing it
>>8698209https://files.catbox.moe/4hwurq.jsontry running your dataset for around 1500-1600 steps with this if it's mostly one outfit. maybe more if there's more
>>8698236open MS paint, select the colour, some brush and remove them yourself then inpaint over it at low denoise
>>8698230>prodigyI never got this shit to work right either I'll give it another whack though, what Loss type do you use or is it whatever?
>>8698279Not him but the main drawbacks of prodigy is that since you can't set the LR, BS 1 will often blow up unet. Also it's very very slow compared to adamW. Still from my tests back on pony it was the best for characters (when BS 1 would succeed).
>>8698284Well I don't have enough memory to fuck with it anyway from what I can tell, it won't run no matter what for some reason. Figured 12gb would be enough.
>>8698286I haven't used prodigy in a bit but I was on 8gb so it should work. Maybe try and see if there's an 8 bit version. Either that or it's some other setting that's exploding your vram.
>>8698136we, normal people, can't fathom the mindset of people who have headmates
>>8698136Who are you talking about?
>up nearly 3k pixiv followers over the last two monthsneat
>>8698321teach me the secrets to goon art
>>8698343bibiciiiiiiiiiiiiiiiiiiiiiiii
>>8698354true people seem to really love bbc stuff kek
:(
>>8698078>cc lin, qingli ye, muted colori mean it's quite a bit better at not making a complete mess and understanding artists, fishnets are still too hard though and it's still slow as FUCK
>>8698378prompting it is fucking hard
yeah it probably won't get much better than thiseuler a cfg++ works best, (emphasis:x) doesn't seem to work. repeating tags has an effect at least.grouping tags within the prompt for girl/male has a pretty large effectfor some reason view between legs/ass is fried into it
>>8698050check "no norm" in settings lil bro
>>8698378>muted color is a proper tagHUH
>>8698279prodigy is mostly a meme, even when the rest of your settings are good>>8698406i still don't get why model development in both llms and txt2img went in the way of "oversized and shitty"well okay it's obviously a vc/proprietary platform money grab but it's still depressing
I wish multiple batches were better for loras. They save so much time.
>finally tried lowering the te lr meme>took longer to converge>looked shittier than the regular lrwaow
>>8698482why do you even train the TE
>>8698484Why not?
>>8698484it's betteryes i tested itmultiple times
>>8698503a surprisingly sensible answer
>>8698510the funniest thing is that it's better even when training without captionsno i don't know what silly quirk of the code does it
>>8695737It's too little. Even tentacles are sparse.
>>8698513Lora training is a meme, there are no rules. Anyone who tells you otherwise is a meme too.
>>8698503this
>>8698568There is a rule, KISS. I tested every single toggle in this fuckin mess and the only thing I actually kept was some memory optimizations, kek.
>>8698568that's not true after doing this for three years and training maybe 10,000 loras i've discovered at least two or threethey all only apply to datasets though
>>8698568there are plenty of basic rules that always apply like GIGO
so this is a "training is a meme" meme time
>>8698639>still better than Tifa spam
>>8698633Honestly nah, at least XL is pretty damn good at generalization even with very few images. I've included some garbage in datasets before, even deliberately, and it bakes out if your params aren't shit.
>>8698644what kind of retarded logic is this? "if you do a decent job then only sabotage it a little bit it's fine?" how about you actually do a good job in the first place
>*babble*
^ retard status?
>>8698654>>8698653
status status?
>>8698738nicedo you have a part 2, on the downswing?
>>8698749I don't have it but I could do it later
>>8698343just do what the great ay*kon wishes he could domake stuff that doesn't look like AI slop to normies and submit it without flagging it
more like belongs in /trash/
I can't believe kanikananika lora bro decided to get himself a vacation like that.
who and what
>>8698738>orc
who did what?
>>8698752>ay*konwho
the most successful AI artist
>>8698867dead namefag, used to shill his patreon here
>>8698738>>8698749>>8698751There, I think it looks a bit odd but good enough I guess>>8698810I am a half-orc in the current dnd campaing that I am playing, perfectly fine self insert :^)
>>8698867>>8695037
>>8698870oh hey, thanksyou got it all in somehow
>>8698872>you got it all in somehowpower of proper usage of tags
>>8698873please elaborate it's actually kind of annoying to control this
>>8698889I need you to be more specific, what are you trying to do?
>>8698893balls deep except actually have it work instead of not
>>8698894just don't use any tag related to penis and use tags related to penetration instead, here is the base gen if you are interested>https://files.catbox.moe/hrgb9r.png
>>8698897that's usually what i try to do but thanks bud
>>8698873how is the gen so clean
>>8698921It's noodlekek, clean gens are his specialty.
>>8698921prompt clean artists (with lora(s) probably)ez
>>8698934this but for real in a not ironic way
>>8698935my posts are always 100% genuine or an obvious joke
>>8698937same here
>>8698934truebuilt-in artists seem to turn noob into a melt machinegood loras are the strat
>>8698939it really depends
>>8698939this sounds more like a skill issue from (you)
>>8698939I use lora (singular, gopro), precisely for the purpose of turning noob into a melt machine.
>>86989422021 called, they want their gen back
>>8698939>meltworks on my machine, sure you're not retarded?
>>8698738>>8698870do a human on elf version of that sometime
>>8698952noelfs are meant for orcs
>>8698952I think I could but it would be a different elf, probably a smaller one so the size difference would be still there
>>8698939I've always seen the opposite, except when I'm overwriting tags noob already knows. Artists it knows are always the cleanest due to the tag.
>>8698237Tried this shit still didn't cook at all it's a little better I dunno if I need to increase LR or what at this point for characters. I'm baking this with nai vpred v1 so not sure if it's vpred issue as well with it.
>>8698995This is why I don't believe in "le special step count." You should bake it until it fries and eyeball each epoch until you find the right one. If you don't want to wait then increase the LR.
>>8699003Yeah step count shittery was memes for me too but like 20 epochs and it kinda works at 16 and that's if I feature prompt like a boomer. I just don't get why shit is so much worse to do then on pony or even older 1.5 shit like noobai just refuses to learn charas in my experiences.
>>8699005What does your dataset look like?
>>8698995noopred and illustrious do need different settings. you at least should be turning off multires noise>>8699003that's around the point where i've stopped seeing improvements but my datasets are simply built different so ymmv
>>8699011>Turn off multires noiseI always have that shit on from before and saw it on in the two .jsons anon's posted earlier what's it do that fucks with noob / illustriou?
>>8699005noob is way better at learning charas than any other model lolit's the styles that's the problemwtf kinda dataset you have
>>8699013it's a feature designed to work with eps that fucks with the noise to make the model at dealing with very bright or very dark images. idk if it does anything specifically bad with vpred but i assume you probably don't want or need it. i just have it at 10/0.2 because it make loras slightly sharper in tests i did earlier
>>8699014Never had to tailor this hard with other charas it just got them before so I dunno if I got to re-learn this shit or whatever. Artist styles usually turn out fine
why can't my style lora do darks anymore? I don't have multires noise on. What causes it to unlearn darkness?
>>8699020>my style lora do darks anymoreYou can't call them that anymore, it's African-American.
>>8699022No the term is darkies, now answer my question.
>>8699013multires noise is noise offset applied multiple times at different "resolutions" which allegedly works better than regular noise offsetnoise offset was used in sdxl 1.0 (and subsequent models until vpred) to make up for shit and piss you get when you use non-ztsnr schedule without any tricksif you use ztsnr schedule you don't need any of that shit
>>8699023I dunno. Are you even training on vpred?That's what's supposed to do darkies
>>8699032On both. On some datasets I get lucky can do dark, dim lighting on vpred and on others it just stops working entirely.
>>8699034I guess you're just overfittingor using shitmixes and exacerbating all the eps mixed in them with the lora
>>8699027Be honest chief I got no clue if I even have ztsnr on or what it is.
>>8699036>I guess you're just overfittingNope since i'm testing every epoch and even the early epochs that learned little can't do it.
>>8699038i dunnyo what i can tell yait just werks for menot like those artists have a single dark pic in them kek
>>8699041Box?
>>8699044https://files.catbox.moe/sw847w.png
>>8698943box pls?
>>8698955don't worry, go for it when you feel like it.
>>8699079https://files.catbox.moe/nhv80i.png
>>8699106>jellytits>tits look like waxdropped
>>8699106Thanks bro!
we're STILL on SDXL??????
it's nai
no we're on neta lumina keep up
let me guess, you need more
Just for 3 more months, sir
How do I mosaic to post on pixiv btw
>>8699149prompt it???
>>8699140neta lumina could very well be the next big thing
>>8699152it's not, it has 8 epochs of real training and 100 epochs of aesthetic finetuning
until i see something complex and not pov cowgirl from neta lumina, it's dead to me
>>8699149literally any image editor besides paint can do it
Unironically it needs to reach NAI's level of complex poses before I'd make the switch from noob.
>prompt "1girl, 2boys, double anal, sandwiched" on 102dcustom>get one dude with twin peniseshuh?
>>86991641.0 2/d/ custom
good way to conveniently sharpen sections? ie I rotated or scaled a section in gimp and just want to resharpen it instead of completely regenerating it.
>>8699157ok turns out its retard proof easy on photoshop good to know
>>8699164use more tags?
>>8699167Not without serious quality loss. You should inpaint it. At low denoise so that it doesn't change too much, then maybe blend the original with the inpaint if you really don't want it changing anything.
>>8699168you dont need to censor thatyes its retarded
>>8698873catbox please?
>>8698870why is it so melted?>>8698932what's so clean about it? looks pretty standard
>>8699304Standard is this >>8699168 though...
>>8699152It could have been if someone finished training it, but right now it's barely at illustrious 0.1 levels while also being several times slower.
>>8698752>make stuff that doesn't look like AI slop to normies ai have never seen a singke image posted itt that comes even close to not looking AI. but then again we are talking normies>>8699308i wouldn't say that is the standard but then again he posted the most basic bitch 1girl, standing, blank background in a very simplistic artist style.basically that is like asking a driver who drives at 10mph how he never had a crash in his life
>>8699152>60s per meltie gen on a top tier gpuyeah i'm sure
>>8699318if it finished training and they released fp4/8 versions it couldve maybe been something
>>8699319except the inevitable part when the finished training is still meltie and you need tuning or shitmixing to make it usable like every other model on recordwe barely got one sdxl tune and it's effectively 8 times smaller
>>8699322Explain why it would be "meltie" (did you mean melty?) if it finished training.
>>8699323kys bot
>>8699328Answer the question Rajesh.
