Previous /sdg/ thread : >>101328700>Beginner UI local installFooocus: https://github.com/lllyasviel/fooocusEasyDiffusion: https://easydiffusion.github.ioMetastable: https://metastable.studio>Local installAutomatic1111: https://github.com/automatic1111/stable-diffusion-webuiComfyUI: https://github.com/comfyanonymous/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 forksSD.Next: https://github.com/vladmandic/automaticAnapnoe UX: https://github.com/anapnoe/stable-diffusion-webui-ux>Run cloud hosted instancehttps://rentry.org/sdg-link#run-cloud-hosted-instance>SD3 info & downloadhttps://rentry.org/sdg-link#sd3https://education.civitai.com/quickstart-guide-to-stable-diffusion-3https://aitracker.art/viewtopic.php?t=57>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://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 news07/09/2024>Tailor3D: Customized 3D Assets Editing and Generation with Dual-Side Imageshttps://tailor3d-2024.github.io>PaintsUndo: A Base Model of Drawing Behaviors in Digital Paintingshttps://lllyasviel.github.io/pages/paints_undo>Video-STaR: Self-Training Enables Video Instruction Tuning with Any Supervisionhttps://orrzohar.github.io/projects/video-star>Compositional Video Generation as Flow Equalizationhttps://adamdad.github.io/vico>Big Tech needs to generate $600 billion to justify AI hardware expenditurehttps://www.techspot.com/news/103699-big-tech-needs-generate-600-billion-annual-revenue.html>Minutes to Seconds: Speeded-up DDPM-based Image Inpainting with Coarse-to-Fine Samplinghttps://github.com/linghuyuhangyuan/M2S>GenArtist: Multimodal LLM as an Agent for Unified Image Generation and Editinghttps://zhenyuw16.github.io/GenArtist_page>AID-AppEAL: Automatic Image Dataset and Algorithm for Content Appeal Enhancement and Assessment Labelinghttps://github.com/SherryXTChen/AID-Appeal>Align and Merge Deep Learning Modelshttps://github.com/shoroi/align-n-merge>UltraEdit: Instruction-based Fine-Grained Image Editing at Scalehttps://ultra-editing.github.io07/08/2024>JeDi: Joint-image Diffusion Models for Finetuning-free Personalized T2I Generationhttps://research.nvidia.com/labs/dir/jedi>PartCraft: Crafting Creative Objects by Partshttps://github.com/kamwoh/partcraft>AnySR: Realizing Image Super-Resolution as Any-Scale, Any-Resourcehttps://github.com/CrispyFeSo4/AnySR>SwarmUI 0.9.1-Beta: Adds PixArt Sigma & SD3 controlnetshttps://github.com/mcmonkeyprojects/SwarmUI/releases/tag/0.9.1-Beta>ComfyUI Subgraph Nodeshttps://github.com/vivax3794/ComfyUI-Sub-Nodes>T2IShield: Defending Against Backdoors on T2I Diffusion Modelshttps://github.com/Robin-WZQ/T2IShield>FDS: Feedback-guided Domain Synthesis with Multi-Source Conditional Diffusion for Domain Generalizationhttps://github.com/Mehrdad-Noori/FDS
>mfw Research news07/09/2024>CrowdMoGen: Zero-Shot Text-Driven Collective Motion Generationhttps://gxyes.github.io/projects/CrowdMoGen.html>The Tug-of-War Between Deepfake Generation and Detectionhttps://arxiv.org/abs/2407.06174>ANOLE: An Open, Autoregressive, Native Large Multimodal Models for Interleaved Image-Text Generationhttps://arxiv.org/abs/2407.06135>Structured Generations: Using Hierarchical Clusters to guide Diffusion Modelshttps://arxiv.org/abs/2407.06124>Sketchy Moment Matching: Toward Fast and Provable Data Selection for Finetuninghttps://arxiv.org/abs/2407.06120>Layered Diffusion Model for One-Shot High Resolution Text-to-Image Synthesishttps://arxiv.org/abs/2407.06079>T2VSafetyBench: Evaluating the Safety of Text-to-Video Generative Modelshttps://arxiv.org/abs/2407.05965>Deciphering the Role of Representation Disentanglement: Investigating Compositional Generalization in CLIP Modelshttps://arxiv.org/abs/2407.05897>HiT-SR: Hierarchical Transformer for Efficient Image Super-Resolutionhttps://arxiv.org/abs/2407.05878>Audio-driven High-resolution Seamless Talking Head Video Editing via StyleGANhttps://arxiv.org/abs/2407.05577>Ada-adapter:Fast Few-shot Style Personlization of Diffusion Model with Pre-trained Image Encoderhttps://arxiv.org/abs/2407.05552>Rethinking Image Skip Connections in StyleGAN2https://arxiv.org/abs/2407.05527>Exploring Phrase-Level Grounding with Text-to-Image Diffusion Modelhttps://arxiv.org/abs/2407.05352>An Improved Method for Personalizing Diffusion Modelshttps://arxiv.org/abs/2407.05312>VisioBlend: Sketch and Stroke-Guided Denoising Diffusion Probabilistic Model for Realistic Image Generationhttps://arxiv.org/abs/2407.05209>MJ-Bench: Is Your Multimodal Reward Model Really a Good Judge for Text-to-Image Generation?https://arxiv.org/abs/2407.04842>Segmentation-Free Guidance for Text-to-Image Diffusion Modelshttps://arxiv.org/abs/2407.04800
I'm going to look up names of dental disorders so I can include them in my prompts for Japanese women. Like people did with "moe fang" on drawings.. (Endearingly not offensively.)
>>101332640I finally bested you. I just had to break the laws of physics
>>101332588Funny how AI picks up on the body language news readers use to convince their viewers of a narrative
>>101332666that prompt just had "slutty anchor" in it so the AI posed them accordingly. you can see the token 'slutty' sneaking into a bunch of the text
>>101332599off to a good start
>>101332988Trish is Irish
Slo thread
>152 images last threadYou broke the matrix
Why does the news update guy not post them in ldg instead since they actually talk about new things there and it's just schizo discord ebins here? Idgi
>>101333585Bevausr this is his stated territory
>>101333585avatarfags are banned in /ldg/
>>101333279nice clean style
>>101333585lmao this nigga thinks this thread is for "discussion"instead of avatartrannies jerking each other off
>>101333644you misread the post saar
>>101333644read it again
>>101333279>>101333480I have a feeling you would like the character Clover's artstyle from Virtue's Last Reward.
>>101333585Why don't you post them?
/sdg/ ded?
Get ready to catch
G'mornin Anons, have a great day!
>>101332653>>101332662nice
Where do you stand on dreambooth vs kohya vs onetrainer?I've trained loras with onetrainer and it is a good process but the results are a bit disappointing. I dug into the metadata for the best lora I've got from civit and it was apparently made with dreambooth. Only recently as well which is surprising given the sentiment is that it is dead.Is it all just image quality and selection or is there something to these methods?
>>101332570What does SD have to do with pedotranime?
>>101334835Looks like a goosebumps cover
>(Kate Bush:1.3)
civitai challenge, level: impossible: Make the website stay up with no bugs or errors for 24h
>>101335658>>101335675>>101335691>and if I only could, I'd be gennin up that hill
Mornin
quiet thread
Good morning, anons! I hope everyone is doing well :]>>101336317Morning!!
>>101336368why you on so early
Damn haven't been here in 3 months. The fuck happened? Yet another schism?
>>101336407I went to bed super early last night hahaha! It's also my day off! :D
>>101336449have you looked into the new Kolors model yet?
>>101336449>I went to bed super early last nightLol...I too went to bed earlier than usual last night. I should make it a habit.
what does that have to do with stable diffusion?
>>101336540I haven't heard of that one yet! Is it on Civitai? I can try it out :]>>101336607Honestly same! It's pretty nice waking up early! You have so much more time in the day hahaI like being up late, but so much more is happening during the day!
>>101336661>I haven't heard of that one yet! Is it on Civitai? I can try it out :]I'm not that anon but you can try it out online here and get the nodes if you like it for comfyhttps://gokaygokay-kolors.hf.spaceNodes: https://github.com/kijai/ComfyUI-KwaiKolorsWrapper
>>101336645Unless anything new or exciting comes or out, or an obvious new user comes in asking questions then these threads will remain the way they are.
>>101336683Thanks so much, anon!! I'll download it now and give it a try!
>>101336683coolwhat are the odds this will ever run on consumer hardware
>>101336719>Unless anything newwdym? theres always news posted every thread.
Is there anything free and online that does good NSFW? It seems like they either have a filter, require payment for the good stuff, or both. Or they're just generic knockoffs.
>>101336800hmm... depends on your fetishesdall-e doesn't see feet as sexual, so it would do that
>>101336778Nobody really reads the news
>>101336800just do it at home
>>101336959But but I want to use my computer for other cooming purposes, so I can coom while I prepare to coom. That won't work if it's occupied generating stuff locally...
>>101336961meant to post the SDXL upscale
>>101336961That looks really nice, anon!Mine is still downloading a bunch of stuff it seems hahaCan't wait to try it!
I really like the detail it adds! Gonna try adding it to my normal workflow :]
>>101336980it's basically a better SDXL base model. a 16ch vae would have been godlike for this. For upscaling I'm still using XL cause finetunes. Also wondering how much a difference prompting in chinese really makes
>>101337099Tranny :]
>>101337221It does seem like a great base model! I love the way it does her hair too hahaYeah XL is working great for the upscale! I wonder if it could go into SD3, been trying to make an XL to SD3 workflow lately with no luck hahaMight try it with this!
>>101333585when they first split, I tried to keep the news posted in both generals. they [ldg] ended up aggressively attacking me for not favoring their side enough and generally made sure that I wasn't welcome there, so I stopped>>101337099gm pw, you're up very early
>>101337325>they [ldg] ended up aggressively attacking me for not favoring their side enough and generally made sure that I wasn't welcome thereYou claimed to be supportive of both sides when in reality everyone knew who you were backing. Stop lying.
>>101337325Good morning, Debo!! :DYeah hahaha I knocked out around 11:30 last nightIt's been a great day so far! How are you?
>>101337325You should stay here schizo
>>101337316SD3 will have problems if you try to pass on the XL or Kolors latent cause of the channel mismatch. Image to SD3 upscale should work just fine though
>>101336978post that on civit before i do
>>101337391Idk if they have a kolors category yet. also don't know if it would belong in hat category since it's upscaled with sdxl
>>101337389Ohhh gotcha! So I guess there's no way to do that then hahaI wonder if it would be possible if someone made a node to plug into "load image" so instead of relying on channels it would just start with the loaded image as an img2img starting point
>>101337477just do an image upscale of your kolors base gen, pipe it into SD3 vae encode and pass the latent to a ksampler
>>101337543I'll give that a shot! Thanks, anon :D
>the power struggle of avatars seeing who replies to who first
>>101337562Hello.
>>101337589>>101337626>>101337638Good morning, Comfy! :]It's great to see you! I hope you're having an amazing day so far!
Forge question:Does anyone know how you can stop it from switching the model in and out of the VRAM after each single gen? It's actually causing it to run slower than a1111 on my machine (4GB VRAM)
>>101337364>How are you?I'm alright, working through my [iced] coffee and trying to stay cool. you just started playing with Kolors? seems like an interesting model, though I think I saw people saying the anatomy kinda sucks>>101337638rare comfy sighting. nice to see you. working on anything exciting today? felt like there was a ton of interesting techy stuff in the news lately
Why does Debo lie so frequently?
I dunno where the sepia is coming from but I kinda dig it
Morning anons
brainpain>N20YJY
Would be interested in partnering with someone on the technical side to launch AI influencers. You'll undoubtedly learn something about storytelling on this journey ;)
>>101338036debo never lies
>>101338021Nice!! I'm drinking some cold brew right now hahaOn my second cup right nowYeah! I like it so far for the added details! It's a pretty good starting point :]The anatomy seems pretty good, upscaling it cleans up some stuff like the fingers hahaThat's such a cool gen!!>>101338097Good morning, Quokkanon! :]It's so great to see you!
>>101338190Forgot to attach an image.
>>101338036Thread schizo
BTW debo is spamming reports again so there's even less of a reason to engage with the sensitive failed DJ NEET
So how's SDXL 3?
>>101338442waiting for them to re-release it in a couple of weeks. In SAI time it could be months from now though
>>101338188is 4chan the right angle for this, Adam? Shouldn't we have this conversation in Colab again?oh wait you spent all my fucking credits already>R0KT0DDMMV0J
>>101338199Hello PW :)
>>101338466Aight, thanks for the info anon.
>>101338468>is 4chan the right angle for thisMany here thoroughly enjoy the identities they've created so it's not much of a stretch to think they'd love to be AI influencers :]
Someone needs to invent a controlnet for eyes that just lets you set iris position, angle, and color makeup arbitrarily.
>>101338551not wrong but your tone is suspicious Adam. Can you prove your humanity by manually posting your captcha too?>AAGWH0
>>101332570So, I improved the RankinBass Lora a LOT.Please download the better version. (I replaced the old model so just Redownloading will work for those that already downloaded.)https://civitai.com/models/226324?modelVersionId=628634
>>101338551To be one... probably not. You're on the right track though friend :)
>>101337325Calling you a dumb faggot for posting useless shit never stopped you before.
>>101338649You caught me o . o
>>101338652Is it good with porn?
>>101338718Yes....very very good. (try it with both 2d and 3d).Although I can't post it in THIS board
>>101338652>870 megsI'm gonna try it, but damn this is going to fry my toaster.
>>101338875Sorry, it is much much much better than the ones that Civit trained for me.(Although it is painful training SDXL with 10GB VRAM)
>>101338992
WOAHall the avatars are here! :0 !!
>>101339026haha LOL :]
>>101337626I took an absence from these threads but am glad to see the fennec upon return.
>>101338652I can actually imagine that picture in motion. Nice.The SDXL one was cool last time I tried it, so I'm looking forward to trying the new version sometime later.Good luck with the pony one, the style seems to take a big hit just like realistic models do when they merge with pony, but the prompt adherence is hard to beat right now without taking that visual/aesthetic hit.
>>101339327Bear Omelette.
>>101339327that omelette dubear needs some Browning desu>RSVAWN>oh no
>>101339339That image IS from the Pony Model.I added some advice on Tard wrangling Pony in the description.>In motion.What if we threw it into LUMA and see what happens? Also here is a more plastic doll version.
>>101339440>That image IS from the Pony Model.To be clear I just meant that one is decent compared to the other pony examples, and the XL ones outdo all the pony ones imo. Not that it's bad either.>What if we threw it into LUMA and see what happens? Could be cool. I wonder if you could use the local animation tools too. It might incidentally work out well considering stop-motion source with low frame rate animation gens.
>>101339012Heya Fran! Great to see you :]
https://youtu.be/wfcQreCtvO4?t=140I want you guys to know this is what I think of when I read "debo"
>>101339499can you try a rare or medium-rare T-bone? can it cut the meat?>J0DKJY
>>101338903Update: Totally worth it.
>>101339563Thank you for being patient and Beta testing.Civit's training is shit compared to my Local training, (even if I don't have the rig for it lol)
>>101339525Sup P dubs, Just going to post a few gens before vanishing again.
>>101339544
In comfyui, is there a way to concat strings with newlines?
>>101339728you can concat strings, but what are newlines supposed to be?
>>101339828A linebreak. Usually you can use \n or sometimes the return character, but seemingly not in any comfyui node I have installed.
>>101339608I gotta leave here pretty soon too hahaGot some stuff to take care of todayIt's always great seeing you though!
>>101339629cute ty>Gemini multimodal prompt with your image as input>90s photograph of a sirloin steak cut into pieces with a tooth pick planted in the steak piece like a flag pole, contained on a round paper plate with ribbed edges on a vinyl table top in a dingy hotel room with a blank screen>XPJSGP
>>101339927not bad, here's actual prompt>a paper plate on a temporary folding table. on the plate is an badly cooked welldone steak cut into strips and a smear of ketchup. in the background of a depressing living room is an old vintage television set. the overall impression of this scene is very depressing and distressing. a toothpick is stuck into one piece
>>101340080
>>101338021Yes, lots of cool stuff coming out. Stability is kill if they don't get their shit together fast.>>101337652Yes I am, things are going pretty well on my end.
it's getting weirder +8>repeatedly forced to verify humanity>DS0RSN
>>101339327
>TYJD4X
>>101334612u2>>101337221Grand wizard
>>101339848You could just press enter...Empty line does nothing.
>pony doesn't understand my "wrapped in cellophane" bondage fetishLife isn't worth living.
>>101341175we have to talk to HR first>KW8HR0
>>101341175try shrink-wrap
>>101341047interesting that the pupil is almost never round but the iris is
>>101341154What I'm trying to do is connect a color picker node to a gradient node. There is one gradient node that supports only two colors, and another that supports an arbitrary number of colors. The latter is more flexible, and I've created a group node that outputs its format, but the gradient node requires colors to be a single string input, with the colors internally separated by new lines. Right now, I can do it manually by copying and pasting (pic related) and entering new lines manually, but if I could concat strings separated by newlines I could just use buttons.>>101341175https://civitai.com/models/555144/plastic-wrapped-ponysdxlDoes this help?
>>101341289this doesn't look denoised
>>101341289If you are already making custom nodes this shouldn't be a problem for you.
>>101341346Yeah, still figuring out how to make this aesthetic work just right. It started life as pic related, and the idea is to get her looking like an abstract impressionist painting plus impasto.>>101341347Group node, not a custom node. It's just other people's custom nodes wired together with string and bubblegum.
>>101341289>https://civitai.com/models/555144/plastic-wrapped-ponysdxlthat's silly
>>101341289>>101341402ooh nice img2img
>>101341416The AI doesn't judge.>>101341460Thanks anon. Hopefully I can tinker with it some more. That is a very good tummy.
>>101341662:)
is there some better facial restoration I can use with the automatic UI? Codeformer changes the face too much and is unwelcomed on faces that dont look like shit but need touching up
>>101341714Why not just use adetailer?
>>101341402I think you need to customize the nodes then.Open up the custom_nodes/example and perhaps find out how concat node does its thing.
>>101341662glad it doesn't
>>101341737I only gen off and on every few weeks or months and I only have basic bitch knowledgeadetailer makes good looking faces, what exactly is it doing? How do I make it resemble the original faces more?
>/^d\w{2,9}_\d+_\.\w{3,4}$/i
>>101341980>what exactly is it doing? 1) Detect face. 2) Upscale face and surrounding background to a nice 512*x resolution3) Denoise just the face at a higher resolution with a custom prompt you write only for the face4) Once it's denoised, downscale the image back to its original size, and paste it in place.Depending on how you detect the face (segmentation model or not) then you can either denoise just the masked face or the entire face+background.>How do I make it resemble the original faces more?Use lower denoises (closer to ~0.3 instead of ~0.4)Reword your adetailer prompt, or use no prompt
>>101341861an iron maiden album cover if ever I've seen one
>>101341958very ethereal
>>101341958Nice!>>101341947Also very Cute
can't enlarge areola without changing the size of breasts and thickness of torso and arms
>>101342064this model does a pretty good job of mixing surreal creativity with realistic elements. very fun so far>>101342113thanks :)
>>101342154What is the model?
>the cycle restarts I wonder how quick it'll be this time.
>>101342182its called pixel alchemyhttps://civitai.com/models/469255/pixel-alchemy
>>101342122Inpaint.
>>101342209oh that's very ironic heh
>>101342498>consider the following
>>101342504indeed
what is a refiner? Can I use the SDXL refiner on any model based on XL or one with enough XL in it?
>>101342548forget the refiner, it was a scam
>>101342548refiner is basically the same as an upscale/hires pass done with a model specifically tuned on high aesthetic scoring. its a strategy thats mostly been discarded in favor of just having the base model produce more aesthetic outputs. you could still use it though and, yes, you can apply a refiner pass to any model from the same lineage. you could even use an interpolator to refine from other lineages if you wanted to
>>101333774I looked her up, definitely some resemblance to adult/teen Pebbles
>>101342703>Virtue's Last Reward>The Japanese title is Zennin shibou desu
free tickets everywhere>0X8TXH
>>101335694>Not scraping the hell of it to get your own personal archive.Ngmi
>>101342800Nice fluffy soft ears :)>>101342835fken Epic
>>101342779Also that shine and reflection on the armor is rly nice
>>101342567thanks.
>>101342800cute sneakers
>>101342884it really nails some finer details but then others get very melty. I tried a bunch of different upscaling approaches but haven't found anything super reliable
>oops>participationCommandclaw.H>D4PTHP
slow threadWell, good for you, I guess you moved on really easily
>>101343016I'm impressed that all the limbs make sense, Nice>>101343017Epic yet again
i miss schizo anon
Don't mess up this time :)
>>101342567Imagine how many people got 6GB of hard drive space back from this post alone
>>101343161nice
bake up, samurai
>>101343098You'll have to settle for just mentally ill anons
UggghhI cooked up a furry gen that's actually pretty tame, but also *slightly* fetishy. Is /trash/ really the only place for it?
>>101343273just risk it? we all wanna see your fetish next thread
>>101343314It's an anthro horse in overalls with two heads, one of which is kissing another one-headed anthro horse while the other head is speaking
>>101343351sounds more like horror content.. I think its fine aslong you dont see genitalia
no bakers?
>>101343383countdown of 300s wait time took my attention span of baking away
Where is debo to bake?
>>101343417>>101343417>>101343417
>>101343094tyThat belly button through the clothes is typical. Try (belly button) in negatives.