Previous /sdg/ thread : >>101785853>Beginner UI local installEasyDiffusion: https://easydiffusion.github.ioFooocus: https://github.com/lllyasviel/fooocusMetastable: https://metastable.studio>Local installAutomatic1111: https://github.com/automatic1111/stable-diffusion-webuiComfyUI: https://github.com/comfyanonymous/ComfyUISD.Next: https://github.com/vladmandic/automaticAMD 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>Run cloud hosted instancehttps://rentry.org/sdg-link#run-cloud-hosted-instance>Try online without registrationsd3: https://huggingface.co/spaces/stabilityai/stable-diffusion-3-mediumtxt2img: https://www.mage.spaceimg2img: https://huggingface.co/spaces/huggingface/diffuse-the-rest>Models, LoRAs & upscalinghttps://civitai.comhttps://huggingface.cohttps://aitracker.arthttps://openmodeldb.info>Black Forest Labs: Fluxhttps://huggingface.co/black-forest-labs/FLUX.1-schnellhttps://huggingface.co/spaces/black-forest-labs/FLUX.1-schnellhttps://comfyanonymous.github.io/ComfyUI_examples/flux>Index of guides and other toolshttps://rentry.org/sdg-linkhttps://rentry.org/rentrysd>View and submit GPU performance datahttps://vladmandic.github.io/sd-extension-system-info/pages/benchmark.htmlhttps://docs.getgrist.com/3mjouqRSdkBY/sdperformance>Share image prompt info4chan removes prompt info from images, share them with the following guide/site...https://rentry.org/hdgcbhttps://catbox.moe>Discord6wUwtcJsr2>Related boards>>>/h/hdg>>>/e/edg>>>/d/ddg>>>/b/degen>>>/vt/vtai>>>/aco/sdg>>>/trash/sdg
>mfw Resource news08/08/2024>ml_mdm - Matryoshka Diffusion Modelshttps://github.com/apple/ml-mdm>TinyVAE Training Repositoryhttps://github.com/XmYx/tinyvae-flux>InstantX / FLUX.1-dev-Controlnet-Canny-alphahttps://huggingface.co/InstantX/FLUX.1-dev-Controlnet-Canny-alpha>Grounded SAM 2: Ground and Track Anything in Videos with Grounding DINO and SAM 2https://github.com/IDEA-Research/Grounded-SAM-2>ComfyUI xlabs Flux ControlNet implementationhttps://github.com/comfyanonymous/ComfyUI/pull/4260>"Forge is under a week of major revision. Backend Rewrite is 90% finished"https://github.com/lllyasviel/stable-diffusion-webui-forge#under-construction>Fast Sprite Decomposition from Animated Graphicshttps://cyberagentailab.github.io/sprite-decompose>Concept Conductor: Orchestrating Multiple Personalized Concepts in T2I Synthesishttps://github.com/Nihukat/Concept-Conductor>D2Styler: Advancing Arbitrary Style Transfer with Discrete Diffusion Methodshttps://github.com/Onkarsus13/D2Styler>ComfyUI-EbSynth: Fast Example-based Image Synthesis and Style Transferhttps://github.com/FuouM/ComfyUI-EbSynth08/07/2024>ControlNet Canny FLUX.1-devhttps://huggingface.co/XLabs-AI/flux-controlnet-canny>Diffusion1B: Dataset of 1.2 billion images generated by diffusion modelshttps://huggingface.co/datasets/bigdata-pw/Diffusion1B>Diffusers v0.30.0: New Pipelines, New Methods, and New Refactorshttps://github.com/huggingface/diffusers/releases/tag/v0.30.0>Diffree: Text-Guided Shape Free Object Inpainting with Diffusion Modelhttps://github.com/OpenGVLab/Diffree>LLaVA-OneVision Easy Visual Task Transferhttps://llava-vl.github.io/blog/2024-08-05-llava-onevision>IP Adapter Instruct: Resolving Ambiguity in Image-based Conditioning using Instruct Promptshttps://unity-research.github.io/IP-Adapter-Instruct.github.io>SiCo: A Size-Controllable Virtual Try-On Approach for Informed Decision-Makinghttps://github.com/SherryXTChen/SiCo
/g/ bros... I need help
>mfw Research news08/08/2024>How Well Can Vision Language Models See Image Details?https://arxiv.org/abs/2408.03940>Lightweight Video Denoising Using a Classic Bayesian Backbonehttps://arxiv.org/abs/2408.03904>Target Prompting for Information Extraction with Vision Language Modelhttps://arxiv.org/abs/2408.03834>Advancing Multimodal Large Language Models with Quantization-Aware Scale Learning for Efficient Adaptationhttps://arxiv.org/abs/2408.03735>Soft-Hard Attention U-Net Model and Benchmark Dataset for Multiscale Image Shadow Removalhttps://arxiv.org/abs/2408.03734>Openstory++: A Large-scale Dataset and Benchmark for Instance-aware Open-domain Visual Storytellinghttps://openstorypp.github.io/>TALE: Training-free Cross-domain Image Composition via Adaptive Latent Manipulation and Energy-guided Optimizationhttps://arxiv.org/abs/2408.03637>A comparative study of generative adversarial networks for image recognition algorithms based on deep learning and traditional methodshttps://arxiv.org/abs/2408.03568>Hybrid diffusion models: combining supervised and generative pretraining for label-efficient fine-tuning of segmentation modelshttps://arxiv.org/abs/2408.03433>Set2Seq Transformer: Learning Permutation Aware Set Representations of Artistic Sequenceshttps://arxiv.org/abs/2408.03404>A Non-negative VAE:the Generalized Gamma Belief Networkhttps://arxiv.org/abs/2408.03388>FastEdit: Fast Text-Guided Single-Image Editing via Semantic-Aware Diffusion Fine-Tuninghttps://arxiv.org/abs/2408.03355>An Empirical Comparison of Video Frame Sampling Methods for Multi-Modal RAG Retrievalhttps://arxiv.org/abs/2408.0334008/07/2024>MDT-A2G: Exploring Masked Diffusion Transformers for Co-Speech Gesture Generationhttps://arxiv.org/abs/2408.03312>DopQ-ViT: Towards Distribution-Friendly and Outlier-Aware Post-Training Quantization for Vision Transformershttps://arxiv.org/abs/2408.03291
>Hello Anon! I hope you are doing well :]
>5'11">>101791958Goes in the unet folder in models not checkpoints
>>101791976Now I get this... at least the message is smaller. Already tried updating the ComfyUI and pytorch.... nothing
get the flux out
>>101792098desu anon it just looks like you haven't got the flux model downloaded in the right folder or aren't using the correct node to load it
>>101792098refresh the ui and reselect it
is civit messed up right now? I can't login
>>101792098Is your model in unet?Follow a simple guide anon, the basics have been explained in a lot of places.
Does anybody know how I can get ControlNet to work with SDXL on Forge? I have the models but the results are fucked up. It works fine on SD 1.5 models (with the appropriate ControlNet model. I'm not using the same for SDXL)
>>101792156Have you tried reading the docs retard kun?
>>101792113>>101792128I tried it in checkpoints folder and afterwards in unet folder, neither worked. I followed an youtube guide https://youtu.be/KTPLOqAMR0s?si=PGhl3AZJnbeOXHRT&t=178 he leaves it at the checkpoints folder>>101792114Same error message
>>101792231flux is not the same as a traditional stable diffusion checkpointhttps://comfyanonymous.github.io/ComfyUI_examples/flux/
>>101792231Follow another guide.I literally put it in /models/unet/Flux/, clicked on the uner loader workflow node and selected it, it's super simple.
>>101792231just follow comfys simple guide and make sure your comfy is up to date, and go from there.https://comfyanonymous.github.io/ComfyUI_examples/flux/>>101792196I still see ppl using sd15 embeddings in their sdxl gens, so there is that.
>>101792018Had to check long and hard to make sure this wasn't a merchant
>>101792196Have some self respect anon, don't do that.It's mostly people bad habits.Flux doesn't understand the weight stuff, and masterpiece is useless since at least SDXL.
We need a guide for the setup for low-tier PCs
>>101791896occam's razor desu, the data simply isn't there in general and you're getting results sometimes with descriptions because of an odd few images with laion trash captions that are descriptionsthey're going to have used what they already had from sai so laion and midjourneyafaik they haven't said "we recaptioned everything with vllm" that's just conjecture and more likely down to T5
>>101792350>more likely down to T5to clarify, that's responding to long prompts because T5 is a llm
>>101792350Occam's razor for me would be for them to have money to get (more) complete datasets.If they could pay for the compute, they could pay for a good dataset.
>>101792267>>101792278I am going to follow the guide, thanks m8
https://github.com/apple/ml-mdmWtf is this for? I'm a brainlet
>>101792402they know where i am :^)
>>101792457That was the wrong one
>>101792477I bet she hates He-Man
>>101792453yeah and they want nothing to do with you for a reason, faggot
>>101792514stay mad, datalet
>>101792477poor monk seems lost his meat
>all that data >no likes >no job
>>101792533if your data was quality it'd be in use, sweetheart
What's the future of local? Is SD3 a deadend?
>>101792618moar models. sd3 is dead.
>>101792618Looks like it. Stable diffusion wants to earn money now, so they'll only release gimped shit to the public and try to make money from their big models that are API only. But those models can't hold a candle to their proprietary competitors, so why would anyone use them?Flux, which was recently released and came out of nowhere seems to be the next big thing. Takes a metric pisston of vram to run properly and people are still figuring out the best way to train loras for it and finetune it so it becomes more useful, but it blows all of stable diffusion's base models out of the water.
>>101792618The future of local are releases by companies/startups with local models understanding natural language but who never saw a penis/vagege/tit nor ever heard of any known character ever made.Then some autist anon tries to add this stuff back with more or less success.
Flux does a Beaver Tail, she passes, a bit hairy but nice.
Wtf I think I discovered something interesting by accident, even if you put nothing in the negative prompt, if you increase the Guidance Negative, the model will respect the style more and won't overcook Miku style to the model!
>>101792661How much vram does Flux need?
>>101792901if you go for fp8 flux (~12gb + 2gb for the resolution) + fp16 clip (~9.6gb) + VAE (2gb spike at the very end)
>>101792930I "only" have 12...
>>101792618short term: better fluxlong term: generative 3D model + simulator + LLM + 2D background, it's essentially the same as the structure of video games.
>>101792292I'm not the one who do ithttps://civitai.com/images/23243276
>>101792820No seriously this is working, look!https://imgsli.com/Mjg1Nzc0
>>101792901>>10179294112 is good enough if you optimize your pc vram / ram usage. but it's in the "barely" range. keep cfg at 1
>>101792974People on civitai always use retarded shizoprompts that do nothing. Nothing new.
>>101792618SD shot themselves on the foot with SD3, and with a cool new kid on the block it's all downhill from here for them
>>101792292>>101793049Can you prove it actually never helps? AI shit is perfect for cargo culting bullshit.
>>101793100Yes, push the number to 10-20 and you won't see any difference.And make the same image seed without "masterpiece".
>>101793100Do you think Pony score tags would be anything but noise to Flux?
why don't models have prompt feedback anyways? why can't it tell me it hasn't been trained on whatever gibberish I just typed in
>>1017919354090 is en route, can't wait to train some cunny loras
>>101793270almost looks like bonbi....
>>101792986ugh, i guess i'll unload the waifu workflow
>>101793375i dont know who that is
>>101793409it's probably for the best you don't/ttg/ was a wild place
in comfyui, what fucking key press keeps making all my seeds fixed, and how the FUCK do i disable this?
>>101792820Got another example>Hatsune Miku, 80's anime drawing stylehttps://imgsli.com/Mjg1Nzg2
>>101793523how do i download the workflow for comfy?
>>101793569Take this anon: https://files.catbox.moe/m9qr6g.png
>>101792820>>101793523show the left side of the spaghetti pls>>101793583is that the same workflow you posted earlier, with teh dynamic thrshholding and the disabled xyz?
>>101793610>is that the same workflow you posted earlier, with teh dynamic thrshholding and the disabled xyz?yes, DynamicThresholding is mendatory if you want to use high CFG + Negative prompt/Guidance
>>101793615tips for realism?the settings as they are on teh workflow tend to overburn even with no neg and high neg guidance
flickr in progress. going through past years of 'explore' first which are curated best photos. then ill grab all the photos from every user of those. after that ill go through the groupsalso getty pretrain set downloading images, and more scraping in progress, can probably get that dropped tomorrowgn
>>101793695can you post a screen of comfy's workflow with the burnt image and the settings so I can see what's wrong with it?
>>101793695Do you use those new settings on DynamicThresholdingFull? The old ones weren't optimal
>>101793710>>101793769it's literaly the exact workflow posted here >>101793583only difference is prompt, and i'm not using anything weird (or quality tags)on this image i changed cfg on ksampler to 2, the previous image was 4, and the one before was 6so threshholding seems to be making it worse vs just using cfg 1 on the ksampler
>>101793775>threshholding seems to be making it worse vs just using cfg 1 on the ksamplerhigher CFG are meant to make the model understand prompts better, but if you don't use DynamicThresholding the image will collapse, it's something made to prevent the insane burning, but doesn't work at 100% yeah, seems to be working better on anime rather than realism
>>101793775>>101793803oh and different w x h of coursemaybe that's whyhere's at cfg 1.5, i'm gnona play with the other settings then
>>101793821>6 fingersflux is trash
Killing N’wahs and outlanders in the name of CHIM
>>101793844Works on my machine>>101793878Based
>>101792820Ok I did a XY plot (X = sampler I just took euler AND Y = GuidanceNeg scale) and... what the fuck, it looks like the sweet spot is at 10... how lucky was I? I was just puting this number randomly, if I decided to go for 7 for example I would've gotten something glitched and I would never consider this viable or look at it with more details... this is INSANEhttps://files.catbox.moe/rpp8u2.jpg
I thought it would be a cool concept to have a heart with extending tree branches sucking the life out of stuff around it. But then I got bored/forgot and started making goblin girls roasting hearts over a campfire like marshmallows instead
Pic related
>>101793997you're in the news >>101791952
>>101794037Man, I wish. Weed and marathoning anime was the best, haven't done it for years. I miss it a lot.
indeed
>>101794200So what do these mean. Is Comfy hard to use? I haven't found a reason to try Comfy yet, just looks more complicated.
>>101794221Yes absolutely nothing is explained for you in the UI no tooltips no nothing, and there are hundreds of nodes with sometimes 10+ different settings for each and you're expected to go out on your own and figure out what they do.
>>101792820Watercolor works so well with that methodhttps://imgsli.com/Mjg1ODI5
>>101794200what do you hate most about comfy rn?>>101794221some stuff is harder, some stuff is easier. its a very different UI which some people jive with and some dont
>>101794221>>101794282for me the problem is just navigating the spaghetti when you just want to switch out one nodeeven in cleanish workflows it's a pain to switch things out and then changing settings and making sure your new node is not afecting other stuff negativelyit's not a fire and forget kind of thing with comfyuii like genning to relax, and today i've spend like 8 hrs fucking around with nodes instead of just genning cute realistic waifu with animals ;_;i'm about to call it a day on this pain in the ass lel
>>101794318I feel ya. hopefully you get a workflow tuned into what you're aiming for soon and don't get caught in perpetual spaghetti hell.
I was trying to get her to take a sippy out the Kool-aid man's head
What are your thoughts about AMD's "Amuse" tool?
>>101794423I think I saw someone once comment about it saying "AMD won" but then never saw anyone else ever say a single thing about it
>>101794423What is it even used for? "Amuse" kinda sounds like a comedian chatbot that helps you while telling you a joke or two.
>>101793797What model is this? It looks so good!
>>101794423>>101794613It has these additions, and it's somewhat good for users with AMD hardware.
>>101794613>AMD issues a C&D for ZLUDA and launch Amuse They're laughing in your face AMD bros
>>101794718wh-what..?
horrors beyond your comprehension
>>101794733now this is art
>>101794493So cute is this also flux?
>>101794781yes but the style is pretty unreliable. it pops into hd half the time
>>101794661flux
Is there no easy way to prompt edit in ComfyUI like we used to do in Auto1111?I mean the [start:end:step] syntax
>>101794580have you tried forge? it seems like it already supports fluxhttps://github.com/lllyasviel/stable-diffusion-webui-forge/discussions/964
>>101794887https://github.com/ltdrdata/ComfyUI-extension-tutorials/blob/Main/ComfyUI-Impact-Pack/tutorial/ImpactWildcard.md
>>101794887>>101794932oh wait you said prompt editinghttps://github.com/asagi4/comfyui-prompt-control/
>>101793992Damn that's cool
>>101793878Catbox?
Why is AMD so slow?
>>101794927i've been wary of updating forge since last time (last week) when i ended up having to reset the venv and download everything again>>101794945is it finally working? i gave up on it months ago when i was giving comfyui it's quarterly "let's try this spaghetti shit again" shot
nite
>>101795184gn
>>101795163just clone another forge a new folder and create a new venv, test shit with that, I used to run both webui and forge
alright g'nite all>poor kitty
>>101795225gn
>>101795224space on the drive is the issue, not the risk of the update itselfalready flux swallowed up another 30 or so gb (model, vae, updating comfy, updating deps, etc)i'll check out the forge update one of these days
flux can't seem to do silly ms paint drawings
>>101795328>flux can't seem to do silly ms paint drawingsuse this method anon, it'll help https://reddit.com/r/StableDiffusion/comments/1enm9og/discovered_by_accident_a_trick_to_make_flux/
>>101795396I'm waiting until it doesn't double gen times. shits already too slow :(
>>101795328
Debo the retard
>>101795409>I'm waiting until it doesn't double gen times. shits already too slow :(maybe our savior will be AD (Adaptive Guidance) >>101794987
>>101795420he's already seen it in the other thread he just sucks at imggen
>>101795414not ms paint but closer than what I've been getting>>101795420reminds me of 1.5 days when people were trying to figure out solutions to all kinds of stuff. hope that strategy (or any other) works out>>101795427>he's everywhere and everyone schizo
coping retard kek
Good to see the containment thread is keeping him here.
I'm in the totally random style dimension
How do I do xy prompt substitution etc in Comfy?
vibe
https://github.com/lllyasviel/stable-diffusion-webui-forge/blob/main/backend/nn/flux.py#L2Lmao, illya the pedo is seething so much at A1111 he put this so they can't use the flux code from forge. What a huge faggot.
>>101795930I love code-based drama
>>101795930good for him, A1111 is trash
You can't get enough of your drama, can you
>>101795930A1111 is free so they can use it.Stupid drama craving sperg
>>101795930isn't this the comfy / forge wars arc? comfy is doing something similar in his new commits
>>101795930Is it even safe to run FLUX on any other UI that isn't comfy?
>>101795970A1111 is far better for experimentation and throughput, Comfy is when you've only optimized conditions and want to refine
>>101796050>Forge* is far better for experimentation and throughput,fixed 4 u :)
>>101796074What even is Forge? First time hearing of it
>>101795930So what does this mean exactly? Will it take longer for Forge to be compatible with FLUX? Am i understanding this right?
>>101795930>restricting use of GPL coderetard
>>101795930>https://github.com/lllyasviel/stable-diffusion-webui-forge/blob/main/backend/nn/flux.py#L2thats actually aimed at comfy, it will be the same when SD3.1 arrives
I'm still using Forge version f0.0.17v1.8.0rc-latest-276-g29be1da7It just werks really fast and doesn't have an autismal UI.Uninterested in Flux until anime girl LORAs are posted for it regularly on Civitai.
>>101796128This but I'm more interested on celebs, Dafne Keen LoRA is a priority!
>>101796194>3DPDYou do you, anon.
>>101792986Do you get the same results if you put your neg guidance back, and instead turn up positive guidance? I'm guessing that negative is a hack. I don't think it was conditioned on negative prompts?
>>101796226No, turning up the positive guidance will never have that effect, we've tried it before. And yeah, negative guidance has this weird effect of removing the model's bias, I found it completely by accident and I'm glad I did lol
>>101796103you can already test flux on forge, anon
>>101796226>I'm guessing that negative is a hack. I don't think it was conditioned on negative prompts?it's basically an empty negative prompt + Negative Guidance at 10, you can see what I mean by looking at this workflow for example:https://files.catbox.moe/xlxd00.png
>>101796248Thx, I'll look into it.
>>101796288I'll check it out when I gen next for sure
>>101796353Challenge: get it to show all six strings and not fuck them up.
>>101796371I tried
>>101796461I guess not showing the thinner strings is better than previous models have done, which is completely fuck them up.
>human body
>Debo cried himself to sleep>pw dilated himself before crying
>>101792820>>101793992https://files.catbox.moe/3bdsif.jpg Ok I did a XY plot between GuidanceNegative (X) and CFG (Y) and... you're not gonna believe this, but the default parameter I've chosen before (CFG 6 + GuidanceNegative 10) seems to be the sweet spot... I think I never got this lucky in my life what the hell?
>>101796774Does this work for anything other than Miku?
>>101796920I have these>>101795978>>101794191>>101795045
>>101795930based desu
Can you keep your thread schizo debo in your containment thread? thanks.
https://civitai.com/models/625042?modelVersionId=706397Ok I finally get what Adaptive Efficiency means, it's basically the same thing as CFG, but when we're reaching the end of the inference, when the pictures doesn't change much anymore, it reverts back to CFG = 1 -> 2x speed increase at that moment. It's cool but... DynamicThresholding doesn't work well with cfg = 1 and the result is this whiteish picture, I wish I could find a way to deactivate DynamicThresholding when the cfg = 1, is there a node that can do conditions and shit?
>>101797348I put the workflow for those interested, if we can fix this we'll get a nice speed improvementhttps://files.catbox.moe/w5jc8e.pngYou need to download the adaptive node to make it work
https://civitai.com/models/633553?modelVersionId=708301the user said it took $1 total to train on an a100 (you can rent these online), so LORA training seems feasible, you only need one character lora for example and you can gen it forever.their post:Sure I usedhttps://github.com/XLabs-AI/x-fluxfor training on a RunPod A100 SXM instance (80GB VRAM, but only 42 utilised with default settings).If you don’t take into account that I wasted time setting up and used way too many steps (10,000 but 2,500 was enough) for the small number (700) of images I had it cost less than $1USD to train.Note I had to convert the output to safetensors using huggingface https://huggingface.co/spaces/safetensors/convert then used this script https://huggingface.co/comfyanonymous/flux_RealismLora_converted_comfyui/blob/main/convert.py to make it compatible with comfy (and everything else).https://github.com/jhc13/taggui for both natural language and tag style captions.https://www.birme.net/ For cropping and resizing.Going to try this with larger datasets since this LoRA wasn’t expensive. Takes about 2.2 hours to do 10k steps (can be improved) if you don’t save checkpoints too often (which adds like an hour with the default settings).
>>101797639test output with "miku hatsune on a tv screen" and the lora in the provided workflow:
they said the instance prompt was anime art of a girl/woman, last image didnt have it: >>101797674
new data who dishttps://huggingface.co/datasets/bigdata-pw/BIGstockimage2M
I made a tutorial to improve the inference speed by 25% if you are using CFG > 1 + DynamicThresholdinghttps://reddit.com/r/StableDiffusion/comments/1enxcek/improve_the_inference_speed_by_25_at_cfg_1_for/
>pedo and schizo den
ldg is laughing at us again
pitchfork reviews dropping later today
>>101798785I liked your data.
>>101798910thanks! much appreciated
>>101798676this kinda works with dynthresh but it's slowwwwwwbut at least it's not making a mess anymore. still looks burned for realism tho
CLITFlickr dropping soon:tm:Contrastive Language Image Training (data from) Flickrunknown size atm because scrape still in progress, target is 50M+ thoughopen to suggestions for the vllm to use for captioning, preferably something fast so it won't take more than a few days on 4xA100 80GB, testing florence-2 atm
>>101799643Who are your favorite posters?
>>101799801Me, myself and I.
>>101799801every poster is my favourite desu
quick retard questionIf I want to render something in 1920x1920 only shit comes out and it takes forever.How do I fix this?Go for a smaller size and use the Hires.fix?Or is there another way to get the AI to render it properly
multi character grid
>>101799902what model and what interface?1024x1024 is guaranteed to always be good, higher res require you do other things yes, like hires fix.
>>101799956webui Im a retard so I dont know that there are other interfacesdifferent modelsI tried people with absolutereality and some wallpapers, but the wallpapers (I love forests) dont work. Always the same trees.
>>101799801The one that posts the blonde fox. Don't care for the poster, I like the fox, she is cute, and I want to fuck her
>>101799999checked, quints don't lie
>>101799994absolutereality says it's a 1.5 model on civitai. I don't even bother with 1.5 models, they're too finicky.Make things easy on yourself and be sure the "Base Model" section says it's SDXL or Pony.
>>101800063>>101799956Woah, what an utter nonsense advice.
>>101800063thanks man, Im an absolute novice with these things.Open to any feedback, thx anons
Ded general
>>101800364she isn't here
>>101800383Hes always here
>>101800445She isn't here tho
>>101800541Ok debo
>>101800364It's actually working, he stays only here except for his mfw bs on /ldg/
After 1.6 hours in FLUX, I present to you.... my masterpiece.Seriously though WTF am I doing wrong? All the other models gave me much better results...prompt: high quality digital painting portrait of a female rogue, she has a ponytail, brown hair, a rudimentary muffler over her shoulders, subtle gold ornaments on her hair and sharp beautiful yellow eyes
it's not the prompt, some of your other settings must be fucked up. Either post a catbox or screenshot of your workflow
>>101800716Who are you talking to
>>101800696flux will randomly blur for no apparent reason. just hit gen again
>>101800696>/sdg/ - Stable Diffusion General ?
>>101800790here have a sd image
It's dead, Jim.
>install forge>computer performance actively affected by just downloading a fucking model with the civit extension>has yet to start downloading as i'm typing this>never had this problem with regular sdui
pitchfork droppedhttps://huggingface.co/datasets/bigdata-pw/Pitchfork
>>101801152use comfyui
Will you divulge your other pseudonym if you ever get an interview or will you lie so they don't find out? I doubt you'll ever get to that point anyway but I'm curious.
>>101801369huh?
>>101800834Good stuff
>>101801415I'm asking hlky who now goes by big data presumably because no one wants to hire him.
>>101801369gonna assume that's aimed at me, it's on all the datasets anyway>>101801430bigdata is just the name of a group that i'm the only member of. i haven't actually applied to that many positions and i have been hired before, i don't think anyone really cares, i'm probably just being filtered by ATS because i don't have a degree and i'm not in the US
>>101801468Ah. Decent cope desu.
>>101801468Can I join your group? You can call me the Formatter.
I swear to god captioning is literally the worst part of this hobby
>>101801488a job related to data or funding would be nice, but i'll keep doing it anyway, data has always been a big interest of mine>>101801513maybe, do you have experience in beverage preparation, specifically coffee
Is FLUX on A1111 or forge easier to use? I'm having a hard time with those "split sigma" things that i still don't understand completely.
>>101801369I could never bridge that gap because of how thoroughly brain-broken my haters are
>>101801343does it generate faster than base sdui? my specs are fine for generating but that shit takes forever
>>101798676thanks anon, I thought you were a schizo but still decided to give the workflow a tryI've seen a non-zero improvement on the speeds with pretty satisfying resultsI kneel>>101801622are you autistic
>>101801662Distracted.
>>101801622its not empty. __init__.py has the node code. and then it links out to the project page for the paper. as far as specific usage w/ flux, you'll have to experiment or wait for anons to figure out what does or doesn't work. I think rn its just a theory that AG will work
>>101801622also disregard anyone with filename delux_obvious plant
>>101801343what the fuck is this ui, this is way more confusing
>>101801101>JimHe's too prideful to admit he lost and let it go. Plus, he has nothing else going on IRL besides getting yelled at by mommy.
>>101801714Got it working, I tend to spazz out when distracted. It's an obvious brain damage issue.The node seems to work at least. It's clearly faster than normal SDXL for me.I don't have the luxury of testing out Flux because I'd need to update my machine.
>>101798676on a second analysis, I've seen a 40% gen time improvement on flux devfantastic stuff
>d*b* containment general
>this thread is still up
>>101801837ikr? jim is almost as bad as dave>>101801859>I don't have the luxury of testing out Flux because I'd need to update my machine.I think that's what's created a fresh interest in AG. there's been a push to find ways to improve guidance and control for flux because it doesn't work with cfg. AG was recently raised as a potential route though I dunno if there's been anything conclusive
>>101801751He uses other filenames too because so many anons filter him. Luckily you can always spot his gens because they are trash.
Is everything alright in here?
>>101802147who is this handsome beast?
>>101801548Didn't ask don't care
Yo
>>101802255Hey yo! How have you been? Haven't seen you in a while.
>>101802147>this is whos replying to you
>>101802271Kek
>>101802147looks like a schizo
Why is this subreddit so schizophrenic?
>Good morning/afternoon Anon! I hope you are all doing well :]
>>101802355>subreddit
anon I'm confused.what's the difference between/sdg/ and /ldg/?which is more professional and can give me the most technical help?
>>101802385/ldg/
>>101802385>which is more professionalneither, /ldg/ is full of complete amateurs that just discovered Dynamic Thresholding and /sdg/ is a cesspit of avatarfags
>>101802355>nogenWhy are you such a nobody? LOL! :]
>>101802385/lmg/
>>101802400who are the good guys?
>>101802355picrel>>101802385there's no difference beyond arbitrary lines drawn between who's willing to post in which general. you can have the same conversations in either general, talk about the same tech, get the same help
>>101802417none, they exist as two sides of the same faggy coin
cope
>>101802417ldg posters come here every day to shit the place up but I guess it is subjective whether you considering that bad behavior
mornin, 2 minutes before noon
Niggerlas and debo like two peas in a pod
>>101802417the good guys don't post anymore, both generals are worthless dogshit.
>>101802478Good morning, i hope you're doing well friend
>>101799481>still looks burned for realism thoI think for realism you don't need to put GuidanceNeg at 10, you can also decrease the AdaptiveGuidance threshold to get something less burned >>101801662>thanks anon, I thought you were a schizo but still decided to give the workflow a trykek >>101801945>on a second analysis, I've seen a 40% gen time improvement on flux devthat much?? what threshold value did you use anon?
>>101802489is your brain/mind the pod
>>101802440>get the same helpWhen's the last time you gave helpful advice that's deeper than surface level
Is it possible that no one here knows what "schizophrenia" means?
>>101802503where can I post pics anonymously with other chill anons?
>>101802440False, anons wanted to get away from blatant rulebreaking, avatarfags, schizos, drama and especially you thread schizo
Morning anonsEars still too big.
>>101802529upto you really, just lurk on both generals for a while and pick your poison.
>>101802478gm. happy friday. whatcha working on today?>>101802518post a gen>>101802529sdg is usually pretty chill when ldg shitposters aren't leaking in>>101802542>anons wanted to get awaythen why do you come here every single day?>>101802553gm. looks kind of like a primitive ancestral quokka. give it some sabertooth fangs, lol
>>101802595requesting this lady with a sidecut/undercut
>>101802542false, pixart fags got salty over being made fun of and wanted a safe space.
You lied to me /sdg/, flux results are always shit (at least to me)
So the new cope is "they are all from /ldg/ shitting up the thread for no reason!" instead of "its a singular schizo anon who shits up the thread since 1.5 years everday!" ?
>>101802641unironic skill issue, you're just retarded.
>>101802644simple answer is it is a group of trolls
>>101802664Nta but yeah they coordinate in the discord
>>101802691The one in this threads OP?
>101802583>post a genSo the answer is "never", okay.
>101802583/sdg/ is chill because most people left for /ldg/ to get away from you
>>101801714>you'll have to experiment or wait for anons to figure out what does or doesn't work. I think rn its just a theory that AG will workI think it's already working with this method -> >>101798676
>>101802615I tried both sidecut and undercut, neither of them is working
>>101802924bummer. a persistent achilles heel of AI
>>101802996do you still own the glasses in >>101803013 ?
>>101803056His glasses are weird pixel ones not the ones in that pic. That's clearly not even him
>>101802641Kek
>>101803056nah he wears minecraft glasses
Quokka wild!
>>101803178>>101803178>>101803178
>>101803172Damn, this nigga wild!
can flux do proper nail polish?
Filling time!
>>101803214idk whart counts as 'proper'
Filled!
>>101802654I get great results with all other models so flux is the retarded one
>>101803389>I can't use this thing, so ITS retarded, not me!cope
>>101803389Ok vramlet
>even comfy left /sdg/ for /ldg/the absolute state of affairs
