Previous /sdg/ thread : >>100182126>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
https://strawpoll.com/wAg3ArVAWy8Answer the avatar question.
What do I need to do with the "concepts" tab in onetrainer? Making a celebrity lora.
>>100185349the results of this poll are going to change EVERYTHING, /sdg/ will never be the same
>>100185349What if I think they're negligible? That's not a choice.
>>100185405Look pretty good, now,where's the paizuri
>>100185405looks pretty goo
>>100185405catbox onegai
>>100185372I think that is where you tell OneTrainer what your training data and prompt(s) are. See pic related, under "Path" select the folder with your images. Regarding prompts, you can either choose a single prompt consisting of instance tag and class tag, or you create a .txt file for each image in your folder aka manual captioning.For a single prompt, create a .txt file and enter the following:(inst_tag) (class_tag)replace (inst_tag) with something random like m1ku or whatever, but the class tag needs to be a word that best describes your dataset, if it's anime you should use 1girl.For manual captioning, select all images in your folder and re-name one of them to something like "c", you will then see that your images are named after "c (1).png", "c (2).png" and so on. Then create a .txt file and keep duplicating with CTRL+C and V until you have as many .txt files as images. Now do the same re-naming thing with your .txt files. If you did it right you should have a folder that has this structure: "c (1).png", "c (1).txt", "c (2).png", "c (2).txt"...That was the easy part, now go through each .txt file and choose a prompt that you think best describes that image.
>>100185490cool
Did catjak actually leave?
I just bought a PSP go
>>100185538no he's in our hearts forever
>>100185530I couldn't for the life of me figure out the prompt part so I'm thankful you explained it. Every image is already captioned with a matching text file so I'll do "from image file name" it sounds like.
>>100185591nice booba
>>1001853492 or 3 are obvious avatars, the rest not really. No option for this either.
>>100185538catjak is ran? ran is catjak? I never put the two together. pic is ran's mom.
>>100185661I only learned this myself 3 or 4 days ago.
>>100185675I always saw people talking about ran but had no idea who ran was even though I'm here like every night. but I know who catjak is.
>>100185639nice.. it tickles the kid in me
>>100185349>has no idea how to structure a sound poll>will still pretend results are conclusive
>>100185661>>100185675its a troll. catjak hasn't been here since 2022.
>>100185731I thought catjak was the one that posted the victorian catgirls, but then there was a poster who used the filename catjak that was a recent poster. They weren't actually the same?
>>100185651who
Why were we gifted this? The gigabytes of weaponslevel coom i generated are of the charts, and I get to do this shit for free? No censorship? In this decade? Wtf, if only textbased AIs were this uncucked aswell
I can't speak math. how my little guy doing?
>>100185865are there some versions of the llama model that are? I never really got into the text gen stuff.
>>100185761Me obviously
>>100185731Well shit. Was Ran the one making weird-ass but sometimes cool img2img wojaks? Whoever that was, was still active when I was around last october/november.
the other vikings made fun of his sippy-stein
>>100185873Tbh i never looked too much into this aswell in the past months, maybe there are good local run models already out there
>>100185894they're always too pussy to say just goes to show they're always talking out they ass
>>100185916no, ran has never made anything cool or interesting. he just makes the same picture for months
>>100185936What kind of pictures?
>>100185872the right one should go down
>>100185930I don't know much about it, but I thought the llama model that leaked was uncensored and I've seen them talk about one called oobabooga. Here, I found it by trying to figure out how it spell it properlyhttps://github.com/oobabooga/text-generation-webui/wiki/LLaMA-model/dc4b59d0fff7861eeb320a8a5bb5560957e15886
>>100185872damnnot goodsmooth loss should go down, or stabilizewhat params are you using? lr, optimizer, optimizer params, scheduler
>>100185916No, I know he's not basedjak anon. That guy sucks.
>>100182842Thanks demongirl anon!
>>100185985>>100186059absolutely tragic. I tried the base XL lora settings.
>>100186087looking at the params, it's fineyou have 200 warmup steps, and the graph is only at 190 steps, it should stabilize after step 200don't know much about XL training, but 0.0003 seems a bit too high for my liking. I usually go with 1e-5
>>100186064Ran continues to be a myth
Anons how do I improve my gens? Workflow should be in image I think? Just issues like little artifacts, looking a bit 'overbaked' etc. Don't know if I should just gen more or I have bad settings for like steps and stuff.
>>100186149Gotta catbox the image, direct uploads get metadata removed by 4chin
>>100186146I guess so...
>>100186149you have to upload it to catbox.moe since 4chan removes all exif data
>>100185873Llama is good but still nowhere near GPT4 level, and especially not runnable locally. The image gen local model scene is incredible right now. Basically better than the commercial models due to customization. And almost as good anyway.
>tfw AI can't tell the difference between Down Syndrome and Fetal Alcohol Syndrome
gentlemen
>>100186170>And almost as good anywayYeah, some anon was posting in the thread last night saying dall-e is better, I went and looked at the thread on /g/ and I really think pony beats it overall. It still might be better at 2 characters interacting, but that's about all.
>>100186232needs more skin showing
>>100185865If you have 24GB of VRAM try TabbyAPI + SillyTavern with this model https://huggingface.co/LoneStriker/Mixtral-8x7B-Instruct-v0.1-LimaRP-ZLoss-3.5bpw-h6-exl2/If you don't have that much then llama.cpp + sillytavern and look for models in GGUF format, those can be partially offloaded to CPU+RAM
>>100186298Needs larger breasts
>>100186504Nice. LK anon would really like this. Haven't seen him in a while
Good morning /sdg/
>>100186513I think he finally managed to buy his own computer and he's been on a week long gooning sesh, genning LK and Azulas
my penis so biggu
>>100186153>>100186166Thanks anons!https://files.catbox.moe/vy2n81.png
>>100186550>seshNigger language model please stop posting
>>100185865You are as blind as you are stupid I suppose. This truly is the decade of a retard.
>>100186504very nice>>100186513>>100186550Yes, but playing Helldivers not gooning, lol
>>100186597>This truly is the decade of a retard.Ironic
>>100186620cool played a bit of the first one. playing system shock remake but got stuck about 5 hours in
do you guys know of a way to keep the prompt data in an image after editing it through GIMP?
>>100186652>keep the prompt data in an image after editing itJust keep the original file as well. If you really want you could probably manually add it back in with an exif editor.
>>100186669>Just keep the original file as well.I do that but the edited ones I used to keep in a separate drive and when I accidentally nuked my HDD I lost 3 years of outputs and all I was left it was my edited folder. I just wanted to retarded proof it even harder cause the usual method clearly isn't powerful enough to contain my stupidity.
>>100186689I've not used gimp in a while but if you've already looked around for options to not remove any exif and not found one then I'm not sure what else can be done. Perhaps there's a plugin that can do that? (not necessarily SD related but just exif in general)
>>100186652Uhh maybe consult computing 101 before asking here.
>>100186715Worthless reply by a faggot
>>100186652I had a python script to transfer prompts made for the /h/ discord channel before it got nuked. no idea where it is. I found this though https://www.old.reddit.com/r/StableDiffusion/comments/13f44zl/script_to_copy_png_info_between_two_png_files/
is /sdg/ healing?
>>100186775From?
>>100186726Oh please go back to your discord group, zoomie.
>>100186823From your lack of ideas and spam.
>>100186835Such anger
Julien, you're getting a dry season. Nobody hires an opioid addict.
I would prefer this insect's wings to look more dynamic, but I have no idea how to prompt that.
>>100186632I also got RE4 and Cyberpunk but I've been spending time in city skylines
>>100186823nogens crying 24/7 about avatars they can just ignore but refuse to do socrybaby zoomer faggots
>>100186900Powerful words, my African-American friend.#BLM #METOO>>100186897
SGkgbXkgZGVhciAvZy9mcmVucyENCkkganVzdCB3YW50IHRvIHRlbGwgeW91IHRoYXQgeW91ciBsaWZlIG1hdHRlciB0byBtZSwgbGV0J3MgZ28gbWVycmlseSB0b2doZXRlciB0byB0aGUgZnV0dXJlIGFuZCBiZXlvbmQh
>>100186910half the avatars have easily filterable names and yet you refuse to do so so you can cry piss and shit yourself here daily wah wah kys faggot enjoy the next year because none of them are going away so keep seething retard
>>100186846we know since at least a month that ani is not julien
>>100186954Base64 isn't that clever. Maybe in African-American circles it is.
>>100186964>>100186974#METOO #FREEJULIEN
>>100186974ESL moment right here.
>>10018697701010010 01101110 01010110 01101010 01100001 01111001 01000010 00110101 01100010 00110011 01010101 01110011 01001001 01000111 01011010 01101000 01011010 00110010 01100100 01110110 01100100 01000001 00111101 00111101
>>100186982see you tomorrow anon, ill be having fun and you will be here complaining again, sad life but hey its your choice
>>100187004Try Navajo instead if you are so clever, retard.
>>100187007ESL
>>10018701901010011 01000111 01101100 01101011 01100100 01011000 01000001 01100111 01010001 01010111 00110101 01101011 01011001 01010011 01000010 01111010 01100001 01010111 01000101 01110100 01100011 00110010 01101100 01101000
>>100186744>>100186709thank you, kings
>>100186964This post could be a pasta because it's so cucked.
>>100187095>a oscilloscope
>>100186910score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up, source_anime, shiny skin, official art, BREAK Laura Kinney, sitting on lap, group sex, multiple guys, faceless man, double penetration, vaginal penetration, anal sex, petite body, swinging breasts, groped, breast grab, blush, orgasm, open mouth, green eyes, long black hair, nude, stockings, jewelry, motion lines, <3, detailed nightclub stage, (huge internal cumshot), age difference, (cum everywhere)
animagine 3.1 is pretty damn good at results even with no loraasuka langley, white blouse, red kilt, school:
>>100187221same prompt except "asuna ichinose from blue archive":
What's the best collection of workflows for comfy? This interface is fucking dreadful and created by some alien.
i've been training character models for a while now and while they look great when im generating close ups, their faces become distorted in full body pictures. the eyes especially turn into shit. is this a limitation with stable diffusion or am i using bad training images? should i be using way more pictures of the person from far away or should i just accept that its the way this works?this is a shitty sample generated while training.
>>100187245Maybe you should take some computing 101 classes first and think about your priorities. If you are unable to find the included examples and workflows you don't have any reason to be even trying image synthesis in the first place.
>>100187007debo
>>100186584I think a first improvement could be to use an upscale model and maybe PAG.
>>100187533nice Meru
sd3 when
>>100187511What is a PAG?Also playing with a new workflow using an upscale and Juggernaut XL as the checkpoint!
>>100187574>What is a PAG?https://github.com/KU-CVLAB/Perturbed-Attention-Guidance
>>100187591Thank you anon! Have a hot goth!
>>100187574It increases prompt adherence a little. I did share the workflow with a comparison output.
>>100185530>>100185582Also important to remember is a good way to train LORAs is to avoid prompting stuff that you want the LORA to handle. Essentially you want to prompt everything BUT the thing your LORA is being trained for (as the "inst_tag" represents that). So if you're training for britney spears and she's wearing a skirt, you would prompt "BritneySpears, skirt, shoes, etc" to prevent the lora from associating her outfit with the BritneySpears tag.
>>100187511
>>100186867Dynamic, as in, in motion? Try to see what you can achieve when prompting for motion/blur on the wings.
>>100187662Dope, looking into it now! Appreciate it the help and examples.
>>100187609Are you blind?
>>100187662Did you guys forget how to use SD?
>>100187662Most of these terms are useless in your case.
>>100187774What do you mean? I'm pretty new to SD and first time using ComfyUI this week so I'm still pretty shit at prompting, any tips are appreciated.
gm
>>100187871post a prompt
>>100187893
>>100186565It’s actually 7.3 inchSo about what you are used to
>>100187921Why don't you express yourself? Only a retard posts the same thing over and over again. You're the perfect example of a person who cries about "fuck artists" because he is completely oblivious to the fact he has never created anything.
>>100187921seems fine to me
pw is cancer
>>100187714No problem. I think it's one of the obvious improvements.>>100187684Looks nice.
>>100187947Huh? No I just have fun with this and have prompted all sorts of weird shit for like TTRPGs and stuff, have no problem with real artists. I just post the same stuff right now to try and limit variables to fix my shitty prompts?
>>100186823Cancer-Stells
>>100187110An*
>>100187947>>100187774Imagine being this much of a faggot. Where are your mindblowing once-in-a-lifetime experience mega-expressive images anon?
>>100188020Why does it always have to be the outliers, expression can be in moderation, pastel colors exist you know
>>100188020Kept you waiting, huh?
>>100186900>>100186964>"I am a nigger", the post
Girls love apples.
Vroom vroom
>>100188067>>100188080Adorbs
>>100188052Not sure what the fuck you're babbling about, but the guy literally said he's new, figuring stuff out and is obviously doing babby's first sexy girl prompts to get a hang of things and you immediately start harping about him trying basic shit and projecting some kind of retarded worldview onto him. You're like the shitty angry art teacher who never amounted to anything and now takes his frustrations out on first-years trying to learn and have fun while not yet being concerned with the nuances of art. IE: A faggot.
>>100187693>Dynamic, as in, in motion?Yes. I have tried your suggestion, along with some other ideas, but it mostly just adds additional wings which are not blurry in the slightest, unfortunately. Probably would be easier to just post process this type of image in gimp or some other editor.
>>100188131Nice writing, love the density.
>>100188137"motion lines" tends to be a strong prompt for adding motion to an image, as it's a fairly common way to convey it, if you haven't tried that yet.
>>100188100
>>100188137>CohesiumIt's also a powerful trigger word, combine this with>Bad Promptin negative.
we have to go deeper
back on the wagon
>>100188223fucking disgusting
>>100188067 >>100188080cute!
>>100188245New medication?
>>100188321What would I prompt to get stuff like this? Vinyl figure? Very cute anon!
https://youtu.be/EWf4uJNZIBY?si=RRsCCMGnKr53UzBUbad hair day is erry day for mig.>>100188348naw, just some financial troubles that are hopefully eased for a bit
>>100188080>>100188176>>100188221fun
>>100188384It sucks, fuck finances!
>>100188368It is easiest if you just try the Pony/SDXL checkpoints:https://civitai.com/user/wangka/models?sort=NewestMost of what gives you a fairly clean image with Pony or SDXL respectively will work with this.
>>100188296low test?
>>100188425high test is wanting to fuck all three bellybuttons?
>>100188412Do you know a good upscaler for Pony? Keep getting weird like chromatic aberration on images after upscaling.
>>100188463Use SD1.5 to upscale pony. Easyfluff because it has artist tag abundance.
>>100188169>>100188204Thanks bros, I've tried all of your suggestions together, combined with some weight increase and also dynamic prompting to try to generate motion effects on top of a more or less normal pose and finally I was able to generate something decent!
>>100188527Cohesium really worked for me.
>>100188503Thank you, is this a thing you can inject into the ComfyUI workflow or like an Img2Img after the fact? Forgive me if this question makes no sense, still learning terms.
The Dreamshaper Turbo model is kinda crazy, takes just a bit over a minute to create this on a 4 year old iPad with 3 steps. Too bad the even faster base turbo model seems to suck, or maybe I suck because I can't make it produce anything non-cartoony.
>>100188552No reply.
>>100188463That shouldn't happen too much with most upscalers. Is it maybe some non-GAN upscale followed by a potentially not well-configured ksampler?
>>100188527Congrats anon, looking good!
>>100188598Maybe? I honestly have no idea, feel like a caveman banging rocks with a stick.
>>100188598The latest is TCD-GOON technology btw. It mixes defragmentation of tensor cores with advanced noise generation without any latent washback whatsoever.
>>100188646This part is probably fine with 4x-Ultrasharp or the NMKD samplers. What's next, a ksampler? Maybe that one is configured a bit too aggressively.
>>100188672Do you mind taking a look at my workflow? Sorry for lewd 2B, think it's embedded? https://files.catbox.moe/e24o6u.png
>>100188650Your input seems fairly useless, at least post gens?
>>100188131I think you (you)d me incorrectly I have nothing about new people genning their first cooming sesh, neither did I tell him offThe comment was about the black/white dichotomy of Fenninger the exact same thing ever, or just doing earthshartingly beautiful stuff every gen It’s enough to put just some of your own soul in there
>>100188691Why don't you go back to your Discord? You're too early, the avatarfag circle jerk starts later on today.
Trying to use comfy is there a way to get a set up workflow? I tried using json file from comfy workflows and I'm missing nodes. Do I manually find those nodes somewhere?
>>100188927Get Comfy Manager or yeah look for the missing stuff manually: https://github.com/ltdrdata/ComfyUI-Manager
>>100188690I do not have / cannot immediately find the first LoRa so I can't actually replicate this but I believe the step count on the second KSampler is too low for the sampler algorithm, maybe try 10-15 steps. It could also be that you want to lower denoise to 0.25 for starters.
>>100188990Thank you very much! I will try this :)
>>100188927Just about everyone installs comfy manager >>100188977 and then uses its functionality to get the missing nodes.
>>100188927just use comfy manager and let it download random untrusted garbage code from the web, nothing can go wrong
>>100189003You are too arrogant.
>>100188552Yes, you can ‘load image’ it like normal
>d*b* trying to stir up shit again
>>100188999And if it's not enough, just further increase the step count to 25-30 and/or try lowering some LoRA strengths to 0.6-0.8.
Maybe it’s an idea to compile a ‘basic how-to of synthography’ and put thát in a pastebin in the op, with all the new people arriving as of late
>>100188977>>100189003>>100189016I found a workflow and missing 10 nodes and now guessing what nodes I need from the manager since the names don't always match up. This is horrible lol My Auto images load without trouble, but I want face detailer and haven't found a decent guide to set it up. I'm trying to give comfy a fair shot.
>>100189063Sorry to keep harassing but if you don't mind sharing info what's the difference between clip strength and model strength on LoRAs? Do I need to put (lora:enterloranamehere:1.0) in the prompt or does the node handle it?
>>100189088I struggled at first, recommend finding a really basic workflow, reverse engineering it, once you know how it all works then slowly adding stuff or customizing to fit your desires.
>>100189069There was one (I still have the link) but some avatarfags deemed it was a "newbie trap" and "outdated". This was coming from the guy who posts furry porn and 3 second "animations".
the figure model is very cool, one of my favorite images i've made
>>100189149You can just say ani, who tf cares anonAnd I still think it’s a valuable resource I’m really effing buisy tonight and tomorrow but I guess I’m gonna put Stells in front of a blackboard one of these days
>>100189217Really effing sharp and detailed anonWell done
>>100189280The mental illness in this general skyrockets when certain posters don't show up
>>100186900Most of it stems from pure envy. We only have 3 avatarfags in this general.
Is it possible to change the default comfy output location or i need to use good ui like stable swarm for this?
>>100189149when was the last update? is the recommended models still shit? does it explain vae?
>>100189371read the command line launch arguments
>>100189268Remember when Ani split the general when he didn't get his way changing the OP?
>>100189376Don't have it at hand. Go to /hsg/ and check their op, they have pretty extensive documentation.
>>100189069What would you put in there? The people who just want the simplest of solutions COULD use a simplified UI.The ones that want to do what people here do - that's a documentation/update effort I'm not sure you are willing to undertake.>>100189088Is "install missing nodes" not working to get most/all of them?>>100189091No problem.>difference between clip strength and model strength on LoRAsIt's because there is CLIP and UNET in SD at a technology level. In most cases you can't easily predict what will do what so usage wise it might nearly as well be slider A / slider B. But technology wise you're influencing a different component of SD.> Do I need to put (lora:enterloranamehere:1.0) in the prompt or does the node handle it?Other UI load the node via <lora:name:strength>, you just do it with a node. There may still be reasons to use (token:1.2) in the prompt with some "token" trained by the lora.
>>100189449Ah so for example if on CivitAI if they say like trigger word: ahegao or whatever I could use that as the 'token'? If something is to be strength 1.0 should I just put both sliders to 1.0? Thanks for the help!
>>100189449Very arrogant post.
>>100189500you're a whiny useless bitch. he's helpful.
>>100189393No, I do remember a certain taur enjoyer doing it, thoAnd I wholly do not give a shit
XL2 when
>>100189449>What would you put in there?Basic stuff, ging from hooking a ksampler to babies first chkp mergeBuisybuisy hallucinating later
>>100189470> Ah so for example if on CivitAI if they say like trigger word: ahegao or whatever I could use that as the 'token'? Exactly. For some LoRa you even have to use the trained tokens before anything noticeable happens. It varies by LoRa tho - others change everything without a token simply by being loaded, or do something very noticeable after being loaded and yet more if a token is supplied.> If something is to be strength 1.0 should I just put both sliders to 1.0?Yes. I also think even on ComfyUI a lot of people use LoRa stackers where they're just combined in one number (until you perhaps decide to toggle advanced mode on an image you really want to try to perfect). BTW often it's also not wrong to also try 0.6-0.8 to see whether letting the base model checkpoint have a bit more influence isn't better after all.
>>100189642You're giving wrong information.
>>100189642Thanks anon, very helpful as a start point for me to try stuff!
>>100189593I think you could do that but I'm not sure it'll attract all that many contributors. It's probably only simple if documented for just one UI or on a specialized webpage that has each alternative in a tab or something... and by now there are quite many.Also the respective projects and user base of course document much of this themselves in some way or another.
>>100189707Damm this look pretty sharp and cohesive,catbox please?
>>100189679No problem. There is a lot of cool stuff to try.
>>100185349>im seven of them
>>100189815Beautiful. Amazing.colors and lighting
>>100189884Thank you! ⸜(｡˃ ᵕ ˂ )⸝It's from https://civitai.com/models/420049
The dress code seem ok to me.
Do any of you have girlfriends
>>100189986Nope, never did
>>100189986sure do, everytime I prompt 1girl;_;
>>100189986have gf, exercise regularly, own house, and work from home
>>100189986used to. never again.
>>100189971luv you're a whores, good quality
>>100190041>exercise regularly, own house, and work from homeNobody asked
>>100189986>Do any of you have girlfriendscertainly none of the 1girlers
>>100189986wife looks like morena, so I gen her a lot
>>100190063triggered
>>100190063how's your life, anon?
>>100190082How so
>>100189815cool filter! Watercolor paper?>>100189845That velvet dress looks so nice on her
>>100189986yeah, i stopped working from home because of her
>>100190131sank yew, i've been going through a velvet phase.. such a unique looking texture
>using 4chanx to randomly swap my filenames between phoneposting, comfy and auto filenamesI'm the schizoest of them all
>>100189416They have less links and it's most deprecated auto shit. have you never used the /sdg/ links?
has anyone tried OneDiff? https://github.com/siliconflow/onediffbeen meaning to try it out but haven't had a chance
>>100190165Yet they still blow out 90% of the posters ITT
>>100190194what do shitty deprecated links have to do with gens?
>>100190156Velvet is so nice! But having a pet makes it near impossible to keep clean.. Have you got any good prompt wording for the Victorian age metal-framed bell-like dresses?
>>100190194you like the bbc spam?
>>100190205Because you're coping over the information when it's allowing them to btfo you on the daily lmao
>>100190182what does it do?
I love koff!
>>100190214esl
>>100190210>Have you got any good prompt wording for the Victorian age metal-framed bell-like dresses?have never tried, but I'd assume something like 'victorian bustle' 'bustle skirt'>>100190227supposedly just speeds up gens
>>100190235Cope first language?
>>100190238Bustle! Ty I heard the word like week ago and forgot what it was.
>>100190259english. how about you darkie?
>>100190259people like you are exactly why I avoid /h/. it really is a shithole now
>>100190131>>100190057She only get eyeliner when she takes off her glasses.
>>100190286>>100190304>seething Don't be insecure
>>100190304this
>>100190339Did you try the (black eyeliner:1.5) feature? You're don't seem to be using an ADetailer for an extra layer after image gen. That helps me a lot with make up and irises.
>>100190136Oof
>>100189762Waifu!Awoo!
>>100189952Flexing your suede to the new people I see? >:)
>>100190406No, I do not use adetailer do you have XL version in mind?
>>100190227>>100189971are you using "big boobs" or so? switch to "busty" instead as it looks like "big" is affecting her face
>>100189986Yes, her name is Stells and she is my catgirl live-in maidShe just flat out refuses to do housework tho…No, anon, I’ve been hurt badly and called a lot of shit by people I considered friends, so I’m out of the game for now since I cannot genuinely love a woman right now
>>100190484I haven't used any SDXL models :( Are you using Comfy or A1111?
>>100190406
>>100189707>"Don't shoot! I was ordered to ritually sacrifice this prisoner, I'm just following commands from the high priest! I mean no harm otherwise! I'm really not into these messy ordeals."
>>100190549>No, anon, I’ve been hurt badly and called a lot of shit by people I considered friends, so I’m out of the game for now since I cannot genuinely love a woman right nowI feel that so deep man. I'm sorry. People suck sometimes
>>100190549>No, anon, I’ve been hurt badly and called a lot of shit by people I considered friends, so I’m out of the game for now since I cannot genuinely love a woman right nowSorry anon
>>100190554you could maybe run your upscale passes with half the cfg
>>100190496good one
>>100190549may you come to genuinely love yourself in the meantime, anon
>>100190575That's with half cfg.
>>100190549Sucking Julien off will only make your condition worse
>>100190602pretty velvet elf :)>>100190554For no ADetailer that face looks rly good. In A1111 you can go into Extensions tab and install adetailer https://github.com/Bing-su/adetailer.git>>100190630Nice!
>>100190667in small scale it looks a lot like the no man's sky logo. Cool gen :)
>>100190549Schubert: D.959 - 2. Andantino, Mitsuko Uchidahttps://www.youtube.com/watch?v=PNu9OxXprq0
>>100190659I'm on comfy, will work a bit harder to get it then.
prickly!
>>100190749who is your favorite sdg poster
>>100190229fuck off julien
>>100190702Going to look for adetailer
>>100190702This face looks really good. Nice round irises. Makeup is on point. teeth are nice and proper. I don't know if ADetailer can do much better.
New Thread>>100190707>>100190707>>100190707
>>100190783>I don't know if ADetailer can do much better.I'm pretty sure that anon also had the comfyui equivalent of adetailer on.
>>100190799I do not, that's raw prompting.
>>100190807same checkpoint as the original prompter too? If so I will give it a try.
>>100190819https://civitai.com/models/112902/dreamshaper-xlwith custom Sigmas14.615, 6.475, 3.861, 2.697, 1.886, 1.396, 0.963, 0.652, 0.399, 0.152, 0.029then upscaler after that.
>>100190131No filter! just the models texture: https://civitai.com/models/420049
