General dedicated to creative use of free and open source text-to-image modelsPrevious /ldg/ bread : >>101405779 Fresh Edition >Beginner UIFooocus: https://github.com/lllyasviel/fooocusEasyDiffusion: https://easydiffusion.github.ioMetastable: https://metastable.studio>Advanced UIAutomatic1111: https://github.com/automatic1111/stable-diffusion-webuiSwarmUI: https://github.com/mcmonkeyprojects/SwarmUIInvokeAI: https://github.com/invoke-ai/InvokeAIComfyUI: https://github.com/comfyanonymous/ComfyUI>Auto1111 forksSD.Next: https://github.com/vladmandic/automaticAnapnoe UX: https://github.com/anapnoe/stable-diffusion-webui-ux>Use a VAE if your images look washed outhttps://rentry.org/sdvae>Model Rankinghttps://imgsys.org/rankings>Models, LoRAs & traininghttps://civitai.comhttps://huggingface.cohttps://aitracker.arthttps://github.com/Nerogar/OneTrainerhttps://github.com/derrian-distro/LoRA_Easy_Training_Scripts>Kolorshttps://gokaygokay-kolors.hf.spaceNodes: https://github.com/kijai/ComfyUI-KwaiKolorsWrapper>Pixart Sigma & Hunyuan DIThttps://huggingface.co/spaces/PixArt-alpha/PixArt-Sigmahttps://huggingface.co/spaces/Tencent-Hunyuan/HunyuanDiTNodes: https://github.com/city96/ComfyUI_ExtraModels>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>Try online without registrationtxt2img: https://www.mage.spaceimg2img: https://huggingface.co/spaces/huggingface/diffuse-the-restsd3: https://huggingface.co/spaces/stabilityai/stable-diffusion-3-medium>Share image prompt infohttps://rentry.org/hdgcbhttps://catbox.moe>Related boards>>>/h/hdg>>>/e/edg>>>/d/ddg>>>/b/degen>>>/vt/vtai>>>/aco/sdg>>>/trash/sdg
blessed thread of frenship
>>101432234 i think since he still wants to do subscription based it wouldnt work out
>>101433885I don't think NSFW will ever work with a traditional monetization strategy because all the payment processors are anti basically everything Pony does. AI stuff will need to be pseudo voluntary or run on crypto.
>>101433582Thank you baker!
buh
Two questions:>1Is Diffusion on CPU memory-bandwidth bound in the same sense as LLMs? Or compute-bound? I'm trying to work out whether OpenCL would grant significant speed-boosts on devices without a dedicated GPU (with VRAM).>2Does any kind of pass-through need to be specified to allow OpenCL to leverage GPU from within a docker container?
>I wrote a userscript for the Kolors generator on huggingface, it converts the .webp output to .png, adds a filename hint for downloaders (subdomain-seed-timestamp_num.png) so it's not just "image.webp" all the time, and injects a json metadata packet of the prompt/seed/sliders/etc into the png comment field of the new images. Would anyone else be interested in something like this?Are you still here?
>>101433885He has a lot of choices now>pixart>hunyuan>Kolorsfuck SAI, I'm glad there's some serious rivals to those cucks now
>>101437145
>>101436717He could put the new model behind voluntary $5+ paywall. I would pay it.>>101435117Psycho-pass police
what killed the threads today
>>101437682Looks like avatar spammers
>>101435417sars pls respond
>Huber Schedule: SNR>Huber Param: 0,1Is there any need to tweak these settings?
Is there any chance to get decent results with a rtx 2070?I tried some llama V3 model yesterday at at most I could get 1 liner replies. Escaping Claude's grasp isn't that easy it seems like. Didn't tinker with the settings in oogabooga yet, however.
>other threads archived alreadythis general lost, pack it up
>>101438826for tagging? theres moondream2 and florence2
any good linux programs for captioning training images? i've noticed that i'm getting better results when training without captions at all, compared to the auto caption function on kohya. but i believe there's some middle ground like only prompting the hairstyle, makeup or clothes that i must be done manually, but doing that by creating a bunch of .txt files with a text editor is too much work.
>>101439832this should work on Linux https://github.com/jhc13/taggui/Using tags means the training takes more time so you want to add few epochs. You could also use caption tag dropout setting to achieve middle ground (this works nicely with style loras).
>>101439869thanks that looks really good. i only train character loras. does tag dropout randomly select tags to train? i was thinking of only tagging things that stick out like weird facial expressions or haircuts. maybe clothing. not 20+ tags that the auto tagging gives me.
>>101439916>does tag dropout randomly select tags to trainyou can set it up different ways. I usually use drop every 2 epoch, which means it drops all the captions every second epoch. I'm not sure if you want to drop captions for character loras, but it's worth testing. I run tagui with wd-swinv2-tagger-v3 and 0.25 probability, it's decent even for realistic photography. It's worth tagging those weird expressions manually if you want to replicate them.
>>101439967>It's worth tagging those weird expressions manually if you want to replicate them.yeah that's what i'm thinking. adding "smiling" would to often just generate a generic smile on top of their faces instead of their actual smile. thanks for the help
>>101439986np dude. South Park characters?
>>101439997nah. doing tyra banks for pony right now
IM IN THE OP COLLAGE TWICE AGAIN WOO
https://replicate.com/p/nq6z3mghs1rm40cgqa08qcbdvg>inequality dating moy ppc indianclaudieyed tim astrbuddipc peas watts search continent trailblazer Wtf SD3M?
Hello anons!
>>101441895maybe "peas" is in the prompt because there's green in the image, heh
>>101441988jannies, no ban pls
>>101442075
>>101442327
>>101442075>>101442327>>101442415why won't you just put a link instead, you can put NFSW in there and it seems accepted by the jannies
>>101442075Nice >>101442451It's censored tho
>>101442460dunno, some people put uncensored pictures on links and they never got deleted so I thought it was an accepted process
>kolors finetuned on midjourney images Grim.
>>101442531how do you know that?
>>101442460>It's censored thowhat that other anon means is to catbox.moe the uncensored then you won't have to worry about being banned.
wowie
why were those images deleted?
>>101443865>>101441988>>101442075>>101442327>>101442415
fuck jannies
>>101443887they were censored thooooo wtf
>>101443898and two of them didnt even have nudity lmao
seems like the janny is on a powertrip or pure jealousy... if youre reading this janny, fuck you
>>101443909>>101443898Probably got mass reported by the usual suspects
>>101444040fuckkkk
>>101443759is this the 1girl printer model that anon shared recently?
>>101444063Yeah, same one
I wonder why chainner feels so much better to use than comfy
>>101433582She really is a perfect girl, Eve, artificial hyperintelligence, flawless wife of Maciej Nowicki
Thoughts on AuraFlow? The creator posts daily updates on twitter, v0.2 may come out on friday according to him. Pic rel is one of the example images that a potential candidate for v0.2 generated (still not finished). I just hope they implement the 32 channel VAE he created on this release