Previous /sdg/ thread : >>101403495 >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 >SD3 info & downloadhttps://rentry.org/sdg-link#sd3 >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 >Animationhttps://rentry.org/AnimAnonhttps://rentry.org/AnimAnon-AnimDiffhttps://rentry.org/AnimAnon-Deforum >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 news07/14/2024>ControlNet-Union-SDXL updated, adding Inpaint and Tilehttps://github.com/xinsir6/ControlNetPlus/tree/main>ComfyBlog July: Versioned Releases, Beta UI, New Modelshttps://blog.comfy.org/new-models-versioned-release-beta-ui-new-models/>danbooru2023-florence2-captionhttps://huggingface.co/datasets/KBlueLeaf/danbooru2023-florence2-caption>Nvidia RTX 50 graphics card family TDPs 'leaked' by Seasonichttps://www.tomshardware.com/pc-components/gpus/nvidia-rtx-50-graphics-card-family-tdps-leaked-by-seasonic>GLSL Nodes for ComfyUIhttps://github.com/patriciogonzalezvivo/comfyui_glslnodes07/13/2024>Kolors + Controlnethttps://github.com/MinusZoneAI/ComfyUI-Kolors-MZ>OpenAI blocked staff from airing security concernshttps://thehill.com/policy/technology/4770116-openai-blocked-staff-from-airing-security-concerns-whistleblowers/>ToonCrafter_with_SketchGuidancehttps://github.com/mattyamonaca/ToonCrafter_with_SketchGuidance>DiffSynth-Studio Adds support for training Kolors LoRAhttps://github.com/modelscope/DiffSynth-Studio/blob/main/examples/train/kolors/train_kolors_lora.py07/12/2024>Progressive Growing for Stable Diffusion Web UIhttps://github.com/rmcc3/experiments-progressive-growing>Video Diffusion Alignment via Reward Gradienthttps://vader-vid.github.io/>MAVIS: Mathematical Visual Instruction Tuninghttps://github.com/ZrrSkywalker/MAVIS>Pallaidium Adds Kwai/Kolors Supporthttps://github.com/tin2tin/Pallaidium#change-log>SEED-Story: Multimodal Long Story Generation with Large Language Model https://github.com/TencentARC/SEED-Story>15M Multimodal Facial Image-Text Dataset https://huggingface.co/datasets/OpenFace-CQUPT/FaceCaption-15M>OpenDiLoCo, open-source implementation of DeepMindās Distributed Low-Communicationhttps://www.primeintellect.ai/blog/opendiloco>CLIPTrase: Explore the Potential of CLIP for Training-Free Open Vocabulary Semantic Segmentationhttps://github.com/leaves162/CLIPtrase
>mfw Research news>07/14/2024>Feature Diversification and Adaptation for Federated Domain Generalizationhttps://arxiv.org/abs/2407.08245>Synchronous Diffusion for Unsupervised Smooth Non-Rigid 3D Shape Matchinghttps://arxiv.org/abs/2407.08244>SoupLM: Model Integration in Large Language and Multi-Modal Modelshttps://arxiv.org/abs/2407.08196>Foundation Model Engineering: Engineering Foundation Models Just as Engineering Softwarehttps://arxiv.org/abs/2407.08176>Multi-Object Hallucination in Vision-Language Modelshttps://arxiv.org/abs/2407.06192>Vision-Language Models under Cultural and Inclusive Considerationshttps://arxiv.org/abs/2407.06177>MobilePortrait: Real-Time One-Shot Neural Head Avatars on Mobile Deviceshttps://arxiv.org/abs/2407.05712>VideoCoT: A Video Chain-of-Thought Dataset with Active Annotation Toolhttps://arxiv.org/abs/2407.05355>CountGD: Multi-Modal Open-World Countinghttps://arxiv.org/abs/2407.04619>AWT: Transferring Vision-Language Models via Augmentation, Weighting, and Transportationhttps://arxiv.org/abs/2407.04603>Neural Garment Dynamics via Manifold-Aware Transformershttps://arxiv.org/abs/2407.06101>Multimodal Machine Learning for Automated Assessment of Attention-Related Processes during Learninghttps://arxiv.org/abs/2407.05803>PICA: Physics-Integrated Clothed Avatarhttps://arxiv.org/abs/2407.05324>Spatio-Temporal Encoding and Decoding-Based Method for Future Human Activity Skeleton Synthesishttps://arxiv.org/abs/2407.05573>FALIP: Visual Prompt as Foveal Attention Boosts CLIP Zero-Shot Performancehttps://arxiv.org/abs/2407.05578>Generalized Low-Rank Matrix Completion Model with Overlapping Group Error Representationhttps://arxiv.org/abs/2407.08517>PredBench: Benchmarking Spatio-Temporal Prediction across Diverse Disciplineshttps://arxiv.org/abs/2407.08418
HELP ATTENTIONIm trying to install stable difussion throught automatic 1111, once I ran run.bat I got this error about failing to run torch. Why is this error? I tried method 1 and not method 2 which said I should instally python, is the error because of that?My GPU is NVIDIA GeForce RTX 3060 Laptop GPU
>>101412353It's been quite a while since i've used A1, but as far as python goes, you gotta DL a specific version and not the latest oneI'm sure someone here will be able to help out though! Sorry, it's been so long haha
>>101412378I have not installed python yet, so I had to install a certain version before installing this?
>>101412334It's okay I accidentally deleted my non-AI lyrics just now but it still made me laughhttps://suno.com/song/2e532b63-57ca-4360-8451-8c284c044cc7
>>101412383Yeah! You're gonna want to download it!I'll look for the guide for ya to find the right version :]>>101412414Heyyy Mouse anon!! :DIt's great to see you again! I'm obsessed with Suno right now!LOL!! This is great!! I should try writing a song by scratch haha that sounds really fun!
>>101412456https://www.youtube.com/watch?v=JTuwovSuJp0i guided myself after this video, I clicked update.bat first then run.bat second, should I have not done that
>>101404056>>101404056a1111
>>101412469Uhmmm honestly I would just scrap everything and start over LOLSo I found the guide for you! You also want python 3.10! :]>https://rentry.org/voldyScroll down and you'll see a link to python! I would read everything tho! If I can figure it out, anyone can! Good luck, anon! :]
Just learned that if you put pussy in the negatives all your gens will be male even if they look female. My friend's dad works for StabilityAI and he told me
Say I like the top half of a pic but want to completely reroll the bottom half. What do I do, inpaint half the pic and pick "latent noise"?
>>101412593Yeah well my dad works for Microsoft! Nyaa nyaa nana boo boo!
>>101412456Yeah it's very fun, sorry if I ignored everyone last night was super drunk. Hope you post more songs though!https://suno.com/song/bfca9dab-5be2-49a4-9692-56647d7eb736Suno seems to be down at the moment and I can't actually remember what I posted. I will probably regret this in the morning.
>>101412216Releasing my latest Lora as a public beta again.And hopefully you guys can help test and get showcase images (with credit if you want).It is (X-Men Evolution Style LORA) [funny enough the only one. All other evolution ones are character based].Please download and give it a shot:https://civitai.com/models/578582?modelVersionId=645290
>>101412620It makes really nice landscapes.
>>101412631That is nice
>>101412631>>101412637Thanks you should give it a shot.
>>101412643
>>101412653>It makes really nice landscapes.
>>101412600Yeah you could inpaint the bottom halfIf you kinda like it but it's a little off, then use low denoise, but if you wanna completely re-roll it then make it high!>>101412615LOOL you were good! I was kinda buzzed myself haha! Speaking of which, I made a fitting song for that too hahahaIt's about going to work still drunk from the night before>https://suno.com/song/c2cb2e52-5260-4543-8f75-dd97f8dba35eI remember it was something about the rallyIt got taken down soon after LOL but I still have it on my bar hahahaI like the style of this one!
>>101412662How'd you get that cool style
>>101412605he also told me SD4 is such a powerful AI that if you type in your own address it will make pictures of your house, you can even ask it to show inside and look at actual people naked and stuff, thats why they couldn't release it to the public its only for top level spies and the president
>>101412693>Make pictures of your houseThat's probably possible now with what we have commonly available.
>>101412680This Lorahttps://civitai.com/models/578582/x-men-evolution-style
>>101412710
I am once again begging for good wildcards of backgrounds/locations. PLEASE
anyway of easily switching to another forge fork in windows? the main one seems to be dead
i am making a wildcard, of random tags i am getting from clip interrogating random pics, but not backgrounds
Aura + SDXL
>>101412850>>101412860Heya Comfy! Great to see you :]I hope you're doing well!
nite, tight, sleep, right, dream of 1girls
>>101413113Goodnight!
>>101413113Goodnight, anon! Sleep well! :]
Any free website to experiment with image generation? Every site I check will ask for a payment but honestly I'm just trying to fuck around with it.
>>101413183right from the ophttps://huggingface.co/spaces/stabilityai/stable-diffusion-3-mediumhere's some othershttps://fal.ai/models/fal-ai/aura-flowhttps://gokaygokay-kolors.hf.space
https://www.youtube.com/watch?v=70H03cv57-oThoughts on this LORA guide? I am a 24 gb god so is there anything that would help me get MORE performance rather than less?
>>101413265Fiddling with settings can help a lot sometimes. Like the sampler can effect it per model, and obviously inference steps.Supposedly SD forge or whatever is faster than 111, but I can't really test it
I'm trying to setup SD on paperspace but can't seem to load any models, has anyone gotten this problem?
>>101413317For example I went from 50s-1m gen times at low inference to a nice and comfy six minutes per image at 30+, but it now actually follows my fucking prompt or dies trying.
>>101413317>>101413367Who are you talking to nigga?Tf are you trying to say?
Has anyone used the google collab SDXL training? I just can't seem to ever get it to train things. The one time where it seems to be doing something it seems to get stuck at building text encoders
Last one from me good night anons
>>101413682Forgot image
>>101413682>>101413688Good night, Quokkanon! Sleep well :]
Good night, anons! Hope to see you all tomorrow night!
>>101413882Take it easy dude!
>>101412353The solution is exactly what it says, install pytorch, follow all the instructions provided to you at the source, don't skip anything, don't skip ahead, step by step>>101412378>>101412456I mean bro how is he getting that error without having Python installed lol
>Using the LORAvs>Making the LORA
based Euler (best for those that converge)based Euler aBased 2M KarrasBased 2S a Karrasfuck DDIMfuck DPM2fuck DPM++ 3MUniPC a cringe
>>101414398ya I agree mostly Euler and DPM++ 2S/2M karras (ancestral or not) are probably the best for XL, in SD15 days DDIM was very valid tho.. whats funy I tried SD3 yesterday and nothing but plain Euler works -_-
i miss schizo anon
why are my images turning out like this? the previews looked fine
>huggingface-cli upload hashes everything before starting to uploadbruv whyhashing all 754gb is going to take like 15 minutes, it could be uploading already
>>101414752looks like you are missing a VAE .. what model are you using?
seafood
Morning
>>101414801cute!
>>101414817realisticvisionv6, doesn't say i need a VAE
G'mornin, have a great day Anons,
>>101414860>realisticvisionv6ya thats weird.. civitai lists the newer ones as with VAE included.. I would try using the standard 1.5 VAE once to check if its the problem? or maybe you have configured your UI to auto vae usage and its doubling down on VAE, that could be a problem to..
>>101414878good morningNice textures. which model?
>>101414398>>101414661Not a regular poster here, but I'll take the opportunity to ask. Is Animagine supposed to get completely fucked by certain samples? It seems I'm only getting good results from it on Euler. Even DPM++2M gives fried images.
what next>ideogram>leonardo>lexica>pixai>playgroundsuggestions welcome desu
>>101415199didnt try animagine yet, but yes certain checkpoints are not compatible with certain samplers and more importantly resolutions
>>101415393>certain checkpoints are not compatible with certain samplersAll right, so I take it this is normal, then. I've been perfectly satisfied with SD1.5 all this time, only picked up XL last month, still adjusting to some of the quirks. Thanks for clearing it up.
hello pwhello quokkahello debohello anihello tranhello comfytroonfuck u julien
>>101415605who are you?
>>101415711They call him...schizo...
>tfw he never posted the catjak dox
j.o. 3+ times a day isnt good. gonna try to n.p.goo dmorning
are there any extensions that hightlight tags and/or upen parenthesis etc for a1111?having different colours for tags would be pretty nice too, even if i'd had to manually apply colour to every tag I use
>>101416850>>101416873I'm willing to bet that UIs like A1111 and ComfyUI don't provide fine-grained control over the text in multiline text boxes
>>101416873no idea if this extension is still good cuz I haven't used a1111 in forever, but I used to use tokenizerhttps://github.com/AUTOMATIC1111/stable-diffusion-webui-tokenizer
>>101416896
>>101416850Anon, please. Think of the advertisers.
>>101413038Hello.
will StabilityAI be able to fix Stable Diffusion 3?i think julien is shit (as a human)
>>101417236I'm mad at you cuz you never linked the comfyblog update here. I had to go find it myself like some kind of nobody
>>101417293>>101417305>samefagging schizo
>>101417305Oh it's someone else doing the blogposts. I'll post it here next time if I remember.
[scraping intensifies]
>numbers>suddenly migraine>dog whines>robots talking>tokenize my identity for the dayHello. Theodosius sed SAI staff sux.>GWPXX2
Morning anons
Is the deepfloyd T5 encoder worth getting more ram for?
>>101417926Cuuuute!
>>101417305you are a nobody to comfy tho
>>101417935no, 16gb model+text encoder nearly fills up my 4090 just by itself and all that to be able to write boomer prompts ..? and then it doesnt even understand "afternoon".. pretty meh
>>101418113ram not VRAM. why aren't you offloading it to RAM?
>>101417926gm>>101417935I've never heard about anyone going out of their way to use it so I doubt its worth anything
>>101412216>When you accidentally set the CFG scale to 0
>>101418140ow it does I guess.. still its 18/24gb VRAM for a 1024x1024 img, and also yea 12ish CPU ram .. my pc aint bad .. 5800x, 4090, 32gb ram, but SD3 with txxl5 brings it to the knees (should get 64gb, but isnt an option for the ITX board I use)
>>101412798how are you wildcarding tags that you get from clip interrogation. are you doing that manually?
>>101418267more like 22GB system ram usage for txxl5 actually
also result comparison: this is CLIP SD3 (11GB RAM usage), and this ..
>>101418448... and this is same prompt on TXXL5 SD3, 22GB ram usage.Both this prompt:>phot of an alien space cat on in a crowd of alien visistors and shoppers at a space market in a future city. there is a night sky and sand storm going on, the space cat says "Good evening!" in a speech bubble, painted by beksinski
(2girls)^2
>>101418496mfw colorblind and cant see the number
>>101418496you should make that a seamless wallpaper pattern
>>101418033kek
>>101418496nice to see you. its been forever. kinda thought you quit sdg
mfw finally went too skinny
>>101418827Still would
>>101418713/g/ has just been annoying with the thread splits and drama so I just don't see the need to be here.I still lurk from time to time to look at pictures.
How are you guys not better at this by now it's been like two years and these pictures still look like shit
>>101418868word..
>>101418903show us your picture prowess
>>101418903show us one of your masterpieces
>>101418827X-Ray works
>>101418919>>101418925oof a little defensive today are we? Don't worry if you keep clicking generate eventually it will spit out a "good" picture and you will feel happy.
>>101418919>>101418925>>101418959nogens
>>101418713still posts in the discord tho
>>101418974he's too butt blasted to join the discord with his friends :[
Does SDXL/pony ONLY respond to booru tags, or does it have some kind of language parser to respond to complex descriptions?
>>101419008pony uses boorushit so it's basically unusable
>>101419008SDXL by default does not respond to booru tags at all (StabilityAI did not use any booru content to train the dataset!) .. Pony on the other hand used 2.4 million booru tagged images to fine tune, so it will heavily react to booru tags, a mix of both will work to on pony models
>>101419008it seems to understand a LOT but I had to crank the inference up fairly high for it to behave
>>101419008>some kind of language parser to respond to complex descriptions?that would be txxl5 on stable diffusion 3 .. not in any SDXL model, it still uses clip which is token based not natural language understanding
Poney riding.
playground are doing around 400k generations per day atm, not bad
>>101418496I tried this but I only got two girls.But the pepe turned out fantastic.
>>101419253pepe crying cum
>>101419273He got it in his eye.Very tragic.
>>101419300if she just had two feet
>>101419037>>101418713very cool
>>101419363these take long enough to gen that 2 feet and a hockeysword is what you get
>>101419419>hockeyswordliterally a gamechanger
>>101419419thanks, I like it anyway! also hockeysword soon Olympic sport?
>>101419369>wrong earrings>vision a mess>outfit wrong>hair thing on wrong sideput some pride into your waifu posting
>>101419784Okay, let's see you do it better then
>>101419784sh.. sh.. shocker! I like anons gens>>101419369>>101419262>>101419300really nice quality
Is this a furry? asking for a friend
1supergirl
>>101420024Morena is that you
any new sd news?
>>101420103I grabbed some pony realism models and got lost for awhile down a rabbit hole that can't be posted to blue boards
>ded general
>>101419843gave it a go, different from resolutions I normally work with
>>101420454
>>101420370My model doesn't understand who she is ;_;
>>101420532even base sd 1.5 knows her
>>101420567sd1.5 knew lots more things than both xl and 3. sai made a concerted effort to strip out people, artists, concepts, etc after they got scared of copyrights and privacy and stuff
>>101420598thank you thread schizo
>>101420631do you think he'll ever take the hint
Imagine ppl would be nice just once..
>>101420631>>101420659you always talking withyourself?
>>101420663you're always nice so the longer you hang around, the higher the ANR [average niceness ratio] will beits all a numbers game
>latching on to the only known posters who respond to him anymore its funny when you think about how long goo anon was gone for
>>101420532Yeah, the further we go from 1.5, the further she gets trained out :( I run a detailer with epicphotogasm and a lora to put her back in
>>101420463fixed hand
>>101420785i hope you are doing well
>>101420877im sure you do. in fact the last week or so has been 'fair'
>>101420785that style is super cool. is the a product of the wildcards you've been working on?
just dont lump me into any of your drama, I am just here to give new ppl help, see cool gens and check on latest tech, not to make friends or enemies or join any discord group or anti group
>>101420944https://files.catbox.moe/3htfui.txtused six of the card in the prompt.i dont claim the wildcard is well made, but i like using things i have 'made'.
>>101420987>>101420987>>101420987
>>101420972yoink! def gonna play around with this later. lots of very interesting tokens in here
>>101420955thats funny because the discord is the reason /sdg/ went to shit
>>101421064twas shit long before that, "anon"
i miss the old /sdg/ ;_;
>>101421361don't miss what has gone, simply revel in that you had the chance to experience it
>>101421064And yet it's still in every OP
How's the hardware reqs on Kolors?
now die in peace thread