Previous /sdg/ thread : >>100178978>Beginner UI local installFooocus: https://github.com/lllyasviel/fooocusEasyDiffusion: https://easydiffusion.github.io>Local installAutomatic1111: https://github.com/automatic1111/stable-diffusion-webuiComfyUI (Node-based): https://rentry.org/comfyuiAMD GPU: https://rentry.org/sdg-link#amd-gpuIntel GPU: https://rentry.org/sdg-link#intel-gpu>Use a VAE if your images look washed outhttps://rentry.org/sdvae>Auto1111 forksForge: https://github.com/lllyasviel/stable-diffusion-webui-forgeAnapnoe UX: https://github.com/anapnoe/stable-diffusion-webui-uxVladmandic: https://github.com/vladmandic/automatic>Run cloud hosted instancehttps://rentry.org/sdg-link#run-cloud-hosted-instance>Try online without registrationtxt2img: https://www.mage.spaceimg2img: https://huggingface.co/spaces/huggingface/diffuse-the-restInpainting: https://huggingface.co/spaces/fffiloni/stable-diffusion-inpaintingpixart: https://huggingface.co/spaces/PixArt-alpha/PixArt-Sigma>Models, LoRAs & embeddingshttps://civitai.comhttps://huggingface.cohttps://rentry.org/embeddings>Animationhttps://rentry.org/AnimAnonhttps://rentry.org/AnimAnon-AnimDiffhttps://rentry.org/AnimAnon-Deforum >SDXL info & downloadhttps://rentry.org/sdg-link#sdxl>Index of guides and other toolshttps://codeberg.org/tekakutli/neuralnomiconhttps://rentry.org/sdg-linkhttps://rentry.org/rentrysd>View and submit GPU performance datahttps://docs.getgrist.com/3mjouqRSdkBY/sdperformancehttps://vladmandic.github.io/sd-extension-system-info/pages/benchmark.html>Share image prompt info4chan removes prompt info from images, share them with the following guide/site...https://rentry.org/hdgcbhttps://catbox.moe>Related boards>>>/h/hdg>>>/e/edg>>>/d/ddg>>>/b/degen>>>/vt/vtai>>>/aco/sdg>>>/trash/sdgOfficial: discord.gg/stablediffusion
>mfw Resource news04/25/2024>ComfyUI implementation of SimDA: Simple Diffusion Adapter for Efficient Video Generationhttps://github.com/ChenHsing/SimDA>Nvidia acquires Israeli AI infrastructure co Run:aihttps://en.globes.co.il/en/article-nvidia-acquires-israeli-ai-infrastructure-co-runai-1001477343>Learning Long-form Video Prior via Generative Pre-Traininghttps://github.com/showlab/Long-form-Video-Prior>GaussianTalker: Real-Time High-Fidelity Talking Head Synthesis with Audio-Driven 3D Gaussian Splattinghttps://github.com/KU-CVLAB/gaussiantalker>The state of AI for hand-drawn animation inbetweeninghttps://yosefk.com/blog/the-state-of-ai-for-hand-drawn-animation-inbetweening.html04/24/2024>Narrative Action Evaluation with Prompt-Guided Multimodal Interactionhttps://github.com/shiyi-zh0408/NAE_CVPR2024>PixArt adds 2K Sigma model and Lora supporthttps://github.com/PixArt-alpha/PixArt-sigma#breaking-news->FlowMap: High-Quality Camera Poses, Intrinsics, and Depth via Gradient Descenthttps://cameronosmith.github.io/flowmap/04/23/2024>HiDiffusion: Unlocking Higher-Resolution Creativity and Efficiency in Pretrained Diffusion Modelshttps://github.com/megvii-research/HiDiffusion>Invoke v4.2.0a2 Adds Regional Controlhttps://github.com/invoke-ai/InvokeAI/releases/tag/v4.2.0a2>AnyPattern: Towards In-context Image Copy Detection https://anypattern.github.io/>SVGEditBench: A Benchmark Dataset for Quantitative Assessment of LLM's SVG Editing Capabilities https://github.com/mti-lab/SVGEditBench>DMesh: A Differentiable Representation for General Mesheshttps://sonsang.github.io/dmesh-project/>PDM-Pure: Effective Purification in One Simple Python Scripthttps://github.com/xavihart/PDM-Pure>IDM-VTON : Improving Diffusion Models for Authentic Virtual Try-on in the Wildhttps://idm-vton.github.io/04/22/2024>MoMA: Multimodal LLM Adapter for Fast Personalized Image Generationhttps://github.com/bytedance/MoMA/tree/main
>mfw Research news04/25/2024>DVF: Advancing Robust and Accurate Fine-Grained Image Retrieval with Retrieval Guidelineshttps://arxiv.org/abs/2404.15771>GLoD: Composing Global Contexts and Local Details in Image Generationhttps://arxiv.org/abs/2404.15447>ID-Aligner: Enhancing Identity-Preserving Text-to-Image Generation with Reward Feedback Learninghttps://idaligner.github.io/>CharacterFactory: Sampling Consistent Characters with GANs for Diffusion Modelshttps://qinghew.github.io/CharacterFactory/>MotionMaster: Training-free Camera Motion Transfer For Video Generationhttps://arxiv.org/abs/2404.15789>Sketch2Human: Deep Human Generation with Disentangled Geometry and Appearance Controlhttps://arxiv.org/abs/2404.15889>PuLID: Pure and Lightning ID Customization via Contrastive Alignmenthttps://arxiv.org/abs/2404.16022>Editable Image Elements for Controllable Synthesishttps://jitengmu.github.io/Editable_Image_Elements/04/24/2024>ID-Animator: Zero-Shot Identity-Preserving Human Video Generationhttps://id-animator.github.io/>From Parts to Whole: A Unified Reference Framework for Controllable Human Image Generationhttps://huanngzh.github.io/Parts2Whole/>Adaptive Mixed-Scale Feature Fusion Network for Blind AI-Generated Image Quality Assessmenthttps://arxiv.org/abs/2404.15163>CutDiffusion: A Simple, Fast, Cheap, and Strong Diffusion Extrapolation Methodhttps://arxiv.org/abs/2404.15141>Taming Diffusion Probabilistic Models for Character Controlhttps://aiganimation.github.io/CAMDM/>Multimodal Large Language Model is a Human-Aligned Annotator for T2I Generationhttps://arxiv.org/abs/2404.15100>Perturbing Attention Gives You More Bang for the Buck: Subtle Imaging Perturbations That Efficiently Fool Customized Modelshttps://arxiv.org/abs/2404.15081>CNN2GNN: How to Bridge CNN with GNNhttps://arxiv.org/abs/2404.14822>Enhancing Prompt Following with Visual Control Through Mask-Guided Diffusionhttps://arxiv.org/abs/2404.14768
gaming
Mecha girls
>>100182237>respect my pronouns
where can a man find some good up to date resources for lora training?
>>100182272Might be too late, my lad.
>>100182273Hair and make up done.
>>100182272a training gui and getting hands on
>>100182264+1
>>100182302>...the cat is blushing too
>>100182338that's a hottie
>>100182338That looks chilly sweet
>>100182351She do be cooking. Also cute witch! I need to get back to making some.
>ask for a guide \ resource on the technology board>receives avatards shitpost of an opinion insteadpottery
>>100182374>was given an appropriate response>still complains classic
>>100182264>>100182346kek>>100182374grow up or just steal the metadata off a civit model and just use those as starting setting
girl drop it to the floor i love the way that booty go
@100182397>grow up>I-it wasn't me, it was my coworkers computer moooooomanother Julien classic
>>100182374I don't know what I expected
More stupid fat whores needed ITT
>>100182426he isn't julien
>>100182272Doesn't exist. best I have gotten was this link https://civitai.com/articles/138/making-a-lora-is-like-baking-a-cake
>>100182452domo
https://www.youtube.com/watch?v=WUajwtw2Esc
>>100182338>>100182366I'm not really good at these.
>>100182126
julien is beans
https://huggingface.co/lodestones/Dense-Rectified-Flow-TransformerFuck me, that furry who trained that one model in the past using v_pred method in the past seems to have just published a new model which is most likely a lower budget SD3 trained from scratch by him.HOW DO WE RUN THIS ANONS?
>>100182527Still pretty cool though!
>>100182302Kaguaii
>>100182606What are you prompting for to get these beauties?
>>100182603I didn't read it...
Sorry for the stupid question, but in Fooocus, what exactly are the differences among "inpaint", "improve detail", and "modify content"? Their documentation of these features is pretty sparse.
>>100182627Haha, these are old gens that I did. Just posting them as I try to work up some new stuff with another demon lady I do buuuut I can give you a catbox for one if you want.
>>100182603literally who
>>100182606Thank you.>>100182627and I also would like to know.
>>100182605AAAAA HI DOGGGYY
>>100182627
>>100182648Catbox would be nice if it is up from last night.
AI doesn't "supercharge your creativity". If you lack imagination you will suck at imggen.
>>100182722Ya, ask a kid at the candy store which ONE candy does he want, and he only can pick one.
>>100182478
>>100182721https://files.catbox.moe/axpw1m.jpgThere you go
>>100182778LOL savage
>>100182722>AI doesn't "supercharge your creativity"using that logic meth addicts should be societies most productive
>>100182778Thanks...
So the guy who trained a model in the past using the v_pred method now trained one using a method that is used in SD3, how do we run it anons?https://huggingface.co/lodestones/Dense-Rectified-Flow-Transformer/
>>100182763astronomical
>>100182815fuck I posted in the wrong tab, my bad for spamming this twice
>>100182722I know this feeling all too well. Sometimes I gotta pause because I run out of scene ideas and want more than just poseposeposepose>>100182812Lol, I just randomly pulled one. Here, have the flame onehttps://files.catbox.moe/lx9kpi.png
Girls with serious expressions are cute.
>>100182904I agree! But sadly mine isn't cute though she does have a 'look'.Either way I approve of the cute armored girlNow I should work on some more red demon girl.
>>100182945It's alright to go for sexy rather than cute if you feel like it.
>>100182346lmao
>>100182437You expected too much, anon.
>>100182842Thanks anon.
>>100182952that foot. How?
>>100183006Now I just need to steal...borrow make up from >>100182706I'll be all set. Because mine is always runny
How come my p*ny models act like there isn't a vae in the latest 1111?
>>100183080set it manually. Vae refresh is trash in A1111.
>>100182842I'm having way too much fun with this one.
>>100183087I set it to automatic, none and the base XL one and they were either grey with a box around the ad detailer area or just a black screen
>>100183126yea, that feels like a system restart due to OOM.
>>100183006No problem, friendo!>>100183031I always had to tell it in the negatives 'running makeup' 'running mascara' when ever I saw that happening. Which was a lot when I gen'd for Goth girls. >>100183122Haha, well that's good! Inspiring is always fun!
Is there a way to convert 1.5 LoRA into Pony?
I have a chance to get an RTX 8000 for $2100.It's got 48GB VRAM, but it's pretty much a slightly-better 2080 TI in performance.Then again, the 5090s coming with probably triple the performance, but likely just 24GB VRAM.
>>100183132Still the same for the j2 model
>>100182627Could you share the prompt/model for that one?
>>100183269Never mind a workaround for some extension messed it up
>>100183142Just need to find a better upscaler that doesn't mess with original image at all.
I'm curious, is it possible to use pony LoRAs on other XL models?
>>100183331Most likely it won't work.
>>100183331no.
question for anons with experience doing SD animation:i have an animated sequence of frames - think something simple like a bouncing ball, or car driving past a tree, or dog picking up a bone, etc. these sequences have arbitrary lengths but for the sake of argument, assume 100 frames.for any given frame, i could export it as a still and input into SD either as image2image or txt2image with a depth controlnet, in order to get an output that is close to the shape/composition of my image, but with a completely different style or content or whatever.pretty basic stuff...here comes the real question:what is the current state of the art way to use my exact sequence of frames as input and get an animated output with the least amount of flickering? i dont want any methods that require me to retime things, or pick out keyframes to blend between or anything like that - i want to take my exact frames and process them as is, applying a SD style or re-imagination of content on top, just like a regular txt2image prompt, but with minimal flickering.at first it thought i could find some control net that takes the current frame, and uses the previous frame as guide, to minimize flickering over time, but i cant get solid answers on whether or not that actually exists...all the tutorials for things like animatediff seems to be based on the premise that you'll just be blending between extracted keyframes, or doing copious retiming, or specifically use a dancing human as the input, or smoothing out the flickering in some 3rd party app.....etcany help here? i just want to take an input sequence, and get an output sequence of equal length and framerate, with SD magic on top and minimal flickering. .....possible?
Is the DPO version of Pony better than the regular one?
>>100183425very nice
>>100183417yes. dpo is magic and it's better to have it baked in rather than not. you want autismix dpo over pony dpo.
>>100182706How do you do make up man, I can't seem to work it out.
>>100183450What is dpo?
>>100183361comfyui, animate diff + depth controlnet maybe some masks too. you could throw ipadapter on there as well for something to guide the style with images. animate diff is good because it normalizes the colors and motion and probably your best bet other than deforum loopbacks but that's probably too flickery or unstable for your needs
>>100183471https://civitai.com/models/242825/dpo-direct-preference-optimization-lora-for-xl-and-15-openrailmagick.
>>100183328I am as basic as they come with upscalers and just run 4x-UltraSharp which does the trick for me always. Buuut maybe I could look for another, I dunno. Once you get set in your ways sometimes it is easy to not drift when it works, haha.Anyway, done for now for what I can post here. Was fun! Glad I spread the good demon girl word, lol
>>100183536Thanks for the workflow.
>>100183560Anytime! Have fun GENing!
>>100183450Can I use pony LoRAs on autism mix or that also wouldn't work?
>>100183586works about 60% of the time.
>>100182272https://github.com/Nerogar/OneTrainer to trainhttps://github.com/jhc13/taggui for captionsPut images and captions in a folder. Use a reasonable learning rate with AdamW. Perfect every time.
>>100183586you can bud
>>100183586autism is a pony derivative. you ~must~ use pony loras on it, regular XL ones wont work. theoretically some loras wont work as well on autism as they will on pony but I have yet to see it actually be an issue.
i will carve this 1girl out of latent space myself if i have to
Anyone else compiling a forbidden dataset from the boorus. Over 1.2 million images downloaded so far.
>>100183643what
>>100183648I want to make Pony for Stable Cascade. Hopefully better.
>>100183600>Use a reasonable learning rate with AdamWthis dude doesn't use prodigy lmao
>>100183677isn't there a shill on /h/ doing the same thing?
>>100183704Prodigy is good if you like to fry the shit out of your Loras. Sometimes I think you people have brain cancer with how saturated and sharp you cook your images. You have the visual taste of a toddler licking sugar.
>>100183723>((((sharp focus))))
>>100183717Maybe, I dunno, I can't be the only person who wants to do it.
>>100183752I'm nervous... idk what kind of research quokka is performing
anyone can help me with my workflow? im trying to img2img but the speed tanks anytime i start the queue, i have 4090
>>100183723>no image>>100183751I think it's a little strange to start just before a new model drops
>>100183783You're just some avatarfag with no taste I don't care what you think.
Would anyone be willing to help me turn this into a proper dataset for training? Or at least tell me if it's possible with the images I have. These are scraped IG pics. https://files.catbox.moe/7vb19g.7z
>>100183792>negen calling a nogen an avatar
>>100183766how to lure tourists into quokka traps
>>100183806Anon I already know you don't post a gen because you're an avatarfag who probably bakes the shit out of his images which will only prove my point.
>>100183807nice one
>>100183817who is it?
>>100183766>>100183816Local hosting of Stable Cascade on 6gb VRAM laptop
>>100183723You just can't use it properly. It's still the best optimizer, and is 4x faster as well.
>>100183799You write captions describing each image and the keyword InstaThotWhore. Such as "InstaThotWhore posing in her pink skirt outside, she looks extra skanky today".
>>100183862Maybe for shitting out your overbaked 1girl anime loras.
>>100183853kek
>>100183677You're very unlikely to succeed given that models like Pony are made by pros/organizations but good luck anyway.
>>100183853>the smoke means its working>the flames means its working even better
>>100183888Yeah it's really hard to put a million images with captions in a directory and click train.
>>100183874No, I hardly ever train anime stuff. Prodigy was built on DadaptAdam, which is also another high quality optimizer that blows AdamW out of the water.
art thief maxxxing
>>100183864Yeah but it's 96 images and they're almost all either 3/4ths or fullbody shots. most datasets I see are 15~25 face images and then 0~5 bodyshots. I also don't know if they pass the quality threshold necessary to be useable. tagging I actually understand fairly well now but shit in = shit out so I'm trying to prevent shit in as best I can
https://civitai.com/models/419979/ssbbw-diapereach day we stray further and further away from God's light
Digging through some of my old gens
>>100183924Not the worst thing I've seen on there. I had to filter the 1girl sex with bugs ones.
>>100182603interesting
>>100183898>Dunning-Kruger: The postI would just try to train a Stable Cascade LoRA in general, with just 20 or so images. And forget about trying to do anything close to PonyDiff, do you know how long it'll take to cache latents for that many images? Let alone the training process itself?
>>100183905>stealing his own artnow this guy's thinking with portals
>>100183776catbox it, I'll try it out
a walking plebbit post kek
>>100183736
>>100183888>models like Pony are made by pros/organizationsholy shit the author is dumb as fuck. anyone can overtune a checkpoint when showered in money. just went to the wrong person
>>100183776#1, tiled encode/decode is slower. #2, upscale by shits the bed in python so use the manual upscale image node. just type in the original size of the latent and *1.5 after and it will calculate the expression
>>100184091Nice. It was supposed to be jennifer connely, I think the upscaler used made it a little blurry
>>100184014
>>100184156I think fat girls with small boobs is objectively the worst body type
>>100184172if you haven't fucked a fat fold you haven't lived life
>>100184156
weird, this lora works with pony but not with autism, I just get gobbledygook
>>100184360It can be like that sometimes
>>100184360Pony is so far from SDXL at this point that most loras aren't compatible.
Good evening, anons! I hope everyone is doing well :]I'm not sticking around, just wanted to apologize about last nightI went out last night and got kinda drunkTotally not an excuse tho! I should have just ignored the negativity as usualAnyways, goodnight anons! I totally paid for it today LOL
>>100184451I like feisty witch tho
>>100184451Look at this fucking blog. I don't get why this long enforced global rule is suddenly forgotten in these threads.
>>100184449But it said it was a pony lora and I thought autism was a pony merge, so I would expect it to possibly not work with XL but work on models with pony dna since it does work with base pony and some other pony merges. I dunno.
>>100184468Most regular posters post gens every night that are similar in style and content, I think it's really debatable if that counts as an avatar. No offense to posters like toony, I like his stuff a lot, but he has a style I can usually tell is his. Is that an avatar?
>>100184494If I deliberately deposit $9999 in the bank to skirt money laundering laws, is it legal?
>>100184468you're pretty mad, huh?
>>100184506So what don't other posters like quokka anon make you respond the same way?
>its only one anon
I'm trying to make a character whose hair is made out of clouds and has completly white skin, could AutismMix DPO give me that?
>>100183956Yes about 4 hours, not everyone is limited by a 1060 like you
>>100184211
>>100184519I hate the quokka too, but getting rid of the most egregious example first to create a precedent is better than distributing my hatred between the four or five obvious examples. That and purple bitch has an annoying pattern of basically blogging while posting their avatar. Discussion about stable diffusion hasn't died everywhere else, but these avatar fags have effectively smothered the discussion here and it shows.The rules me be enforced to fix it.
>>100184528Anybody?
>>100184568drop me your prompt let me try.
>>100184528here is an attempt with pixart
>>100183905Pretty good result here. I haven't had much success with doing this to finish pics. It does well on simple stuff but fucks up when trying to complete more complex pieces. Still waiting for perfect img2img to autocomplete my doodles to my imaginations content. What at have now is pretty amazing and useful nonetheless
>>100184360
>>100184550My point was I think his gens are different enough you could argue it's not really avatar posting, avatar posting is that schizo on /biz/ that attached the same exact image to every post and he did get banned, repeatedly, for that. But okay, you do admit you're singling one poster out out of a dozen and you've created some justification for that so I guess my point is moot.
>>100184550I don't mind mild avatarfagging what annoys me is the personalityfagging, trying to differentiate yourself from everyone super hard and form a name for yourself here. Like forcing yourself as a meme.
>>100184468the "it's not an avatar" cope is ridiculous when there are 190+ pages examples of them posting their slop in the archives purple schizo truly living up to our expectations
why are nogens always such little crybaby faggots
@100184614idk why are you?
>>100184592they call me an avatar too, even though I've proven to be one of the most varied posters. "avatar" is just a slur they throw at people whom they recognize and dont like for other reasons
>>100184575Completely white skin, cloud hair made out of clouds, red eyes, big puffy cloud hair, hair over one eye, thick, 1girl, solo, wearing knitted sweater, simple background, arms up.
>>100184591very nice remix anon
Tourist here. What's the consensus on SD3?
I accidentally downloaded sdxl with vae included instead of the base sdxl. Does that matter for lora training?
dead on arrival
>>100184655Something like this?
>>100184614some of them are nogens because they can't gen (too poor) or suck at genning (so have nothing to post). they makes them resentful towards other posters. on the other hand, some of them ARE posters and go anon so as to hide their gens away from their shitty behavior/personality. they know what they're doing and don't want to take ownership of it. >>100184677we haven't gotten to play with it. a version of sd3 was exposed through APIs but SAI said it was and older/badder one. we're not gonna get a good feel for it until we can run it local. rn most of the evidence we have suggests it might suck
>>100184651Enough people hate ai gen that I don't think we need to in-fight with each other as well, but I guess it's an unavoidable thing in any "community."
>>100184700Alright thanks
>>100184700>on the other hand, some of them ARE posters and go anon so as to hide their gens away from their shitty behavior/personality. they know what they're doing and don't want to take ownership of it.>If you don't avatarfag you're a chudschizo mental gymnastics
>>100184677It's not out yet but apparently SAI put out a very underpowered version of it on their paid API.Doom posters are in full swing but I'm optimistic and I hope i can run it.
>avatarfagging is... LE GOOD
>>100183704>>100183723I'm the guy who originally asked for advice, I've been using prodigy and I think it's fine, although it does look a little overbaked, but i'm willing to accept that for faster resultsThe reason I originally asked for a resource is mainly because occasionally my loras will descend into producing noise in just a few epochs, and I can't find a common denominator.
>>100184530Can't tell if trolling or genuinely delusionalMaybe train an uncensored SD4 on your single consumer card while you are at it
>>100184752Again, you are making assumptions because you're a poorfag
>>100184709most of us agree and are friendly with one another, but there's some people who are thoroughly convinced that 4chan is exclusively for hatred and slurs. some people think sdg should gatekeep instead of help. some people think devs should be bullied out of the general. you can't really appeal to these people because their values are not aligned with supporting a community
>>100184694Maybe something a bit more cartoony.
>>100184747Your learning rate or repeats are too high. You should opt for 1 repeat per epoch and aim for like 50-100 or so epochs
>>100184787>supporting a community>some people think devs should be bullied out of the generalare you talking about picrel and how it can't stop talking about other devs for clout?
>>100184805I'm not certain it's the config that's the issue since most of the time the config works well (OneTrainer's SDXL defaults). I'm thinking it's either something else like hardware or user error/retardation
>>100183959Nice design anon I gave her a weapon and made her smug
>>100184847lol I dig it
>>100184607Yeah, it absolutely is an avatar. Anyone who has seen purple witch once can describe it, + or - the hat. Just because stable diffusion allows them to produce variations on that motif doesn't exclude it from being an identifying avatar to go along with whatever they're posting about in the moment.These avatar posters are literally using the excuse that I'm too poor and thus jealous of them as a deflection tactic. This shit needs to be called out and exorcised from the threads.
>>100184839Depending on your hardware it should only take a few minutes to run some trials and tweak the parameters, it's not one size fits all. Even people Loras, I've had some celebrities that took way more epochs to train.
>>100184770Now that you realized I was right, I am actually curious to see a good SC LoRA because I think Stable Cascade still has some potential. Can you post your training config? Kohya_ss, OneTrainer, or even Diffusers cmd args, doesn't matter I can work with all of them.
>>100182126Post prompt and model
>>100184882Your assumption is that you think I care what you think. You're just a poorfag crab
>>100184872.threads the from exorcised and out called be to needs shit This.tactic deflection a as them of jealous thus and poor too m'I that excuse the using literally are posters avatar These.moment the in about posting re'they whatever with along go to avatar identifying an being from it exclude t'doesn motif that on variations produce to them allows diffusion stable because Just .hat the - or + ,it describe can once witch purple seen has who Anyone .avatar an is absolutely it ,Yeah
>>100184788https://civitai.com/models/196786I've been trying it with this but other than the one I've posted I haven't had a success.
>>100184891hol up, I'm starting to get on board with this thread exorcism idea suddenly....... o_o
Remember the era of VHS/Digital artifacts being used as effects in video games? How long until we see the era of SD artifacts as effects?
>>100184788>https://files.catbox.moe/5vsw22.pngGood luck m8
>>100184961use the catbox image to get workflow
>>100184941>How long until we see the era of SD artifacts as effects?not soon enough
>>100184787It's got nothing to do with being rude, avatar/namefag posting has always been hated on 4chan because it goes against the idea of anonymous, original content (even if this is often not the case). This isn't twitter/discord, so trying to establish a presence though tripcode or avatars will always be met with a hostile reaction on almost every board as it is going against a tradition.
>>100185104I genuinely hope you find another thread to backseat mod in instead of obsessing over this one. You cause orders of magnitude more damage than all the people you complain about combined.
>>100184961>>100184979Thx
>>100185104playing self-appointed hall monitor over a random rule no one cares about isn't more important than fostering a helpful and productive community. but like I said, your values are not aligned with that.
>>100184787Some people have expressed negative opinions about voldy in the past as well though. I won't say who they are... Again, I'd rather there not be fighting and I'm not trying to start conflict, but I don't know if by devs you mean on both sides because objectively it has been. And stuff like that doesn't contribute positively to the threads either.
>the avatar cries out in pain as he shitposts
if pastebin were still in OP we would not have this problem
>>100185187nice
Can someone catbox a good comfyui image? I just got it tonight and holy shit trying to set up adetailer is like rocket science
>>100185189Thank you (ꈍᴗꈍ)it is from a finetune I uploaded today, https://civitai.com/models/420049/atilessence
>>100185157Avatar fags are a extremely vocal and hated minority. Just because you don't see the hatred boiling over at all hours doesn't mean it isn't there, simmering every day.The fact Avar fags come out of the blue to defend the practice without being specifically pointed out is very telling about how aware they are of their practices.
>>100185212oooo, this looks very good. saving for future use. hope it works with more than just 1girl!>>100185222>Avatar fags are a extremely vocal and hated minority.wrong
Who has a screenshot of the strawpoll?
>>100185140>so mentally ill he immediately feels insulted and accuses me of being the imaginary boogeyman in his head>>100185157I'm not trying to moderate anything, you must be confusing me with someone because I have nothing against you or anyone else here. I was just proving an explanation as to why there are trolls.
>>100185229It does!
https://strawpoll.com/wAg3ArVAWy8Just wanna feel out how people feel about this avatar question.
>>100185267
>>100185280>strawpoll!!11 cringe fag
Can SD3 do feet?
Ok I'm too stupid to use comfy jfc
>>100185291can i do your feet
>ranfag enters depressive stage of his bipolar disorder after he loses the split threads war>gens get ignored>has a nogen melty declaring he will be leaving /sdg/ for goodrinse & repeat with this sperg
>>100185288>mad because the last one proved you wrong >>100185322who?
>>100185291Pony seems pretty good at them, sometimes an extra toe, but overall good
>>100185299Download the Efficiency Nodes (and potentially RGThree). With the Efficient Loader, Efficient KSampler (and possibly Power Prompt) you'll save yourself a lot of headache.
Next>>100185213>>100185213>>100185213
>>100183776run img2img before upscale>>100184110Tiled is slightly slower but that beats your PC freezing whenever you encode/decode anything larger than 2048x2048
>>100183652Underrated pose
>>100183174train new pony lora on 1.5 lora images
>>100183600and OneTrainer works on any setup?Even ROCm/ZLUDA fags with 8GB of RAM on AYYMD?
>>100184451Your okay pw have a good one
>>100184506Technically, yesWhich is the best kind of yes
>>100185118Novel comfy gen
>>100185187>>100185189Agreed, the cig is actually correctly lit
>>100185222>insults minority>acts surprised when minority reacts Like baked clay shapes
>>100185291I like it better when they are dirtyGood picture anon
>>100182842Could you please catbox this one?
>>100183282masterpiece, (ultra detailed background, delicate pattern, intricate detail), (highly detailed, fine details), best quality, (tan:1.3) teenage slim marionette-cyborg-woman sitting on a sofa, wearing her long hair-as-a-dress, pale blue eyes, wide hips, slim waist,( long blond hair:1.5), (dystopian future skyscraper luxury apartment :1.7), pink sunset, (alien plants:1.2), golden neon silk furniture, (piles of treasure:1.5),( tinted crystal glass curtains:1.5),Negative prompt: (holding:2), (worst quality:2), (low quality:2), (normal quality:2),Steps: 30, Sampler: DPM++ 3M SDE Karras, CFG scale: 7, Seed: 3706965838, Size: 512x768, Model hash: e978db4e56, Model: realspice_v20, VAE hash: 735e4c3a44, VAE: vae-ft-mse-840000-ema-pruned.safetensors, Denoising strength: 0.4, Clip skip: 2, ADetailer model: face_yolov8n.pt, ADetailer prompt: "young girl face, (pale blue eyes:1.3), (black eyeliner:1.2), blue eye shadow, (blushing:1.3), shy,", ADetailer confidence: 0.64, ADetailer mask min ratio: 0.003, ADetailer dilate erode: 4, ADetailer mask blur: 4, ADetailer denoising strength: 0.4, ADetailer inpaint only masked: True, ADetailer inpaint padding: 32, ADetailer version: 24.1.2, Hires upscale: 2, Hires steps: 10, Hires upscaler: R-ESRGAN 4x+ Anime6B, Dynamic thresholding enabled: True, Mimic scale: 7, Separate Feature Channels: True, Scaling Startpoint: MEAN, Variability Measure: AD, Interpolate Phi: 1, Threshold percentile: 100, Mimic mode: Half Cosine Up, Mimic scale minimum: 5, CFG mode: Half Cosine Up, CFG scale minimum: 5, Version: v1.7.0
>>100182778Thank you!>>100183031For Eyes I go with eye color in prompt and addADetailer prompt: "young girl face, (pale blue eyes:1.3), (black eyeliner:1.2), blue eye shadow, (blushing:1.3), shy,">>100183457Awesome gen! Try making mascara, eyeliner or makeup more important in prompt Example: (narrowed pale blue eyes:1.1),(black eyeliner:1.5), Ctrl+arrow up raises the prompt parts value between , symbols
>>100182351>>100182365>>100182627thread faves, ty