Previous /sdg/ thread : >>101234327>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 news07/02/2024>16ch-VAE: Open source 16ch VAE reproduction for SD3https://huggingface.co/AuraDiffusion/16ch-vae>DiffIR2VR-Zero: Zero-Shot Video Restoration with Diffusion-based Image Restoration Models https://jimmycv07.github.io/DiffIR2VR_web/>MimicMotion wrapper for ComfyUIhttps://github.com/kijai/ComfyUI-MimicMotionWrapper>E.T. the Exceptional Trajectories: Text-to-camera-trajectory generation with character awareness https://www.lix.polytechnique.fr/vista/projects/2024_et_courant/>FoleyCrafter: Bring Silent Videos to Life with Lifelike and Synchronized Soundshttps://foleycrafter.github.io/>FORA: Fast-Forward Caching in Diffusion Transformer Acceleration https://github.com/prathebaselva/FORA>Diffusion Forcing: Next-token Prediction Meets Full-Sequence Diffusionhttps://boyuan.space/diffusion-forcing/>RunwayML: Gen-3 Alpha Text to Video is now available to everyone.https://x.com/runwayml/status/180782239641546768607/01/24>Segment Anything without Supervisionhttps://github.com/frank-xwang/UnSAM>MimicMotion: High-Quality Human Motion Video Generation with Confidence-aware Pose Guidancehttps://tencent.github.io/MimicMotion/>PopAlign: Population-Level Alignment for Fair Text-to-Image Generationhttps://github.com/jacklishufan/PopAlignSDXL>B-LoRA for Kohya-SShttps://github.com/ThereforeGames/blora_for_kohya>Nearly half of US firms using AI say goal is to cut staffing costshttps://www.baka.com.au/world/north-america/nearly-half-of-us-firms-using-ai-say-goal-is-to-cut-staffing-costs-20240629-p5jpsl.html06/30/2024>HunyuanDiT-v1.2https://huggingface.co/Tencent-Hunyuan/HunyuanDiT-v1.2>Hunyuan-Captioner https://huggingface.co/Tencent-Hunyuan/HunyuanCaptioner>MotionClone: Training-Free Motion Cloning for Controllable Video Generationhttps://github.com/Bujiazi/MotionClone>SDFX, no-code platform to build and share AI apps, goes open sourcehttps://github.com/sdfxai/sdfx
why is sd3 in the OP
>mfw Research news07/02/24>Improving Diffusion Inverse Problem Solving with Decoupled Noise Annealinghttps://arxiv.org/abs/2407.01521>MIA-Bench: Towards Better Instruction Following Evaluation of Multimodal LLMshttps://arxiv.org/abs/2407.01509>Expressive and Generalizable Low-rank Adaptation for Large Models via Slow Cascaded Learninghttps://arxiv.org/abs/2407.01491>FastCLIP: A Suite of Optimization Techniques to Accelerate CLIP Training with Limited Resourceshttps://arxiv.org/abs/2407.01445>StyleShot: A SnapShot on Any Stylehttps://styleshot.github.io/>TransferAttn: Transferable-guided Attention Is All You Need for Video Domain Adaptationhttps://arxiv.org/abs/2407.01375>Restyling Unsupervised Concept Based Interpretable Networks with Generative Models https://jayneelparekh.github.io/VisCoIN_project_page/>Unaligning Everything: Or Aligning Any Text to Any Image in Multimodal Modelshttps://arxiv.org/abs/2407.01157>Evaluation of Text-to-Video Generation Models: A Dynamics Perspectivehttps://arxiv.org/abs/2407.01094>Blind Inversion using Latent Diffusion Priorshttps://arxiv.org/abs/2407.01027>Cross-Modal Attention Alignment Network with Auxiliary Text Description for zero-shot sketch-based image retrievalhttps://arxiv.org/abs/2407.00979>InstantStyle-Plus: Style Transfer with Content-Preserving in Text-to-Image Generationhttps://arxiv.org/abs/2407.00788>Diffusion Models and Representation Learning: A Surveyhttps://arxiv.org/abs/2407.00783>LLM4GEN: Leveraging Semantic Representation of LLMs for Text-to-Image Generationhttps://arxiv.org/abs/2407.00737>Consistency Purification: Effective and Efficient Diffusion Purification towards Certified Robustnesshttps://arxiv.org/abs/2407.00623>GenderBias-VL: Benchmarking Gender Bias in Vision Language Models via Counterfactual Probinghttps://arxiv.org/abs/2407.00600>Toward a Diffusion-Based Generalist for Dense Vision Taskshttps://arxiv.org/abs/2407.00503
>mfw MORE Research news>The Factuality Tax of Diversity-Intervened Text-to-Image Generation: Benchmark and Fact-Augmented Interventionhttps://arxiv.org/abs/2407.00377>SVG: 3D Stereoscopic Video Generation via Denoising Frame Matrixhttps://daipengwa.github.io/SVG_ProjectPage/>OccFusion: Rendering Occluded Humans with Generative Diffusion Priorshttps://arxiv.org/abs/2407.00316>Prompt Refinement with Image Pivot for Text-to-Image Generationhttps://arxiv.org/abs/2407.00247>Transformer-based Image and Video Inpainting: Current Challenges and Future Directionshttps://arxiv.org/abs/2407.00226>The impact of model size on catastrophic forgetting in Online Continual Learninghttps://arxiv.org/abs/2407.00176>Analyzing Quality, Bias, and Performance in Text-to-Image Generative Modelshttps://arxiv.org/abs/2407.00138>DaBiT: Depth and Blur informed Transformer for Joint Refocusing and Super-Resolutionhttps://arxiv.org/abs/2407.01230>Multimodal Conditional 3D Face Geometry Generationhttps://arxiv.org/abs/2407.01074>Human-like object concept representations emerge naturally in multimodal large language modelshttps://arxiv.org/abs/2407.01067>An Expectation-Maximization Algorithm for Training Clean Diffusion Models from Corrupted Observationshttps://arxiv.org/abs/2407.01014>Unveiling Glitches: A Deep Dive into Image Encoding Bugs within CLIPhttps://arxiv.org/abs/2407.00592>Investigating and Mitigating the Multimodal Hallucination Snowballing in Large Vision-Language Modelshttps://arxiv.org/abs/2407.00569>From Local Concepts to Universals: Evaluating the Multicultural Understanding of Vision-Language Modelshttps://arxiv.org/abs/2407.00263>GalLoP: Learning Global and Local Prompts for Vision-Language Modelshttps://arxiv.org/abs/2407.01400>GaussianStego: Embedding Invisible Information within Generative 3D Gaussian Splattinghttps://gaussian-stego.github.io/
Cursed thread.
>>101245726big boy
first time here?
>>101246017who, me?
Graveyard thread
>>101246036as per the culture of this thread you must attach an image so i know who im talking to and can thus reply accordingly
>>101246102Blame trani
>>101246235what did she do?
The baker likes Kemono friends, hmm?Or maybe just Alpaca.Well, I can understand either way.
>debo is crying >Verification not required
>>101246267>she
>>101246267>be SAI employee>post shota>get called pedo>spend entire week sliding and calling others pedo/ranfag
>spam begins again
Walah just like that, it's over
>>101246471/sdg/ has really changed at this moment
>>101245506Rouge the Bat Pony Lora update is out!https://civitai.com/models/188114?modelVersionId=614956Tell me what you think!Now onto Blaze the Cat!
>>101246291I know the feeling
I've been out of the loop since SdXL released. Is SD3 the best base model now? There seems to be a lot more competition so Im trying to get an idea of where to start testing again
>>101245846Kind of nice and this one too >>101245647I wonder if it's possible to something like this. Maybe I'll try it myself today, not quite sure yet I suppose.
>>101247131ranfag sweating rn
>>101247800sd3 has been a pretty weak entry overall. it really flounders with anatomy and a lot of prompt adherence. the general opinion on sd3 has been pretty low.for newer models however, pixart-sigma performs really well. some would say its a step up from sdxl but the community around it is much smaller so the pace of fine-tunes and supporting work has been pretty slow.>>101247877glad you like them! (impressionist) has been a pretty potent token for these, I think
>>101247916>ranfagWho? haha but for real who?I'm working on updating my Blaze Lora to Pony standards now as well. Any suggestions?
>>101247935I will try out something, I find it difficult outside some specific areas.Sometimes even using "impasto" was not that powerful. But of course it takes lot of tries and so on.I feel like lot of SD15 models are better at this than SDXL.
>>101247944he's just our resident sonic furfag gooner
>>101247974lol. Then he would love my Loras.They are better than other Sonic loras because:The biggest issue with other Sonic based Loras is that it makes it too "anthro / furry" and not enough like Sonic characters and their unique anatomy.And this keeps the soul and unique aspects of Sonic characters well.Here are two examples.https://civitai.com/models/166801/cream-the-rabbit
>>101248027I'll pray for you tonight
>>101248027https://civitai.com/models/124563/sage-sonic-frontiersI'll post one more Sage after this one. I just like Purified Sage playing baseball>>101248043Thanks, I need it.
>impressionist technique, oil painting, impressionist,>portrait of Lord Nelson, pink background, impasto, color splashes,Using AlbedoXL, I think I need to add more keywords such as Cohesium etc.I think at least Albedo tends to favor photographic representation more than artistic because it has been lobotomized somewhat. Might try something else too.
>>101248065
>>101248081Sorry one more. I just love this image.
Vanilla sdxl looks almost identical.
Well I guess it is time to try by artists next.
Most models are "tuned" to generate images like this"...
Does anyone remember the name of that slavic model \ arch? - Wanna see if knows the concept of a Ochipok
>>101248610>OchipokIt should be in terms of any model. Or 'headscarf' or whatever.
Yeah okay I guess I am not that good of a prompteur. The more I mangle and if I want to use'impressionist oil painting by Claude Monet' it won't do anything but even look worse.Using Van Gogh actually gave his portrait.This could be because of TCD scheduler and lora. It's probably great for instagram bullshit but for something else not so much.
>>101248600incredible insight, anon
>>101248881Thank you! I'm writing a paper right now.
>>101248637>headscarfThank you, I had forgotten about this. Close enough to what I would like
>>101248934I'll peer review it
>>101249160Maybe you could also find a lora. But prompting "xxx traditional dress" should get closer results I guess. It is not perfect.
Closest I got, going back to original SDXL setup without any lora bullshit and using vanilla sdxl.>(impressionist oil painting:1.3), (messy oil painting:1.2),>impressionist technique, oil painting, impressionist, portrait of Lord Nelson, pink background, impasto, color splashes, Cohesium,
just in case you were wondering: i hate trani
>>101249337>>98663414
where pw
>>101249586You can download the catboxes but Debo would never ask this and with your bad formatting.You have stolen my images too but it is so easy to call you out because of your 90 IQ.
>>101249608>Debo would never ask thissure I would. I think about pw every tuesday, esp on slow ones like today where he'd be a more-than-welcome energy injection
>>101249633Of course I think about PW too but I do that in complete sentences.
>>101249608>>101249633also, "where X" isn't bad formatting, its a common meme often depicting a monke pondering on the absence of something
some transparent pixel art :)
>>101249649I'm sorry, I was too paranoid and gatekeeping myself. All good though.
>>101249667its understandable. who wouldn't want to be debo? gotta keep an eye out for imposters
>>101249664gotta keep it as png dawg
>>101249695doesn't work on 4chan anyway or at least the pngs are too big, kept the raws as png
>>101249704test I guess
>>101249681Do you have any professional advice how could I mess up these images?>(impressionist oil painting:1.3), (messy oil painting:1.2),>impressionist technique, oil painting, impressionist, portrait of Lord Nelson, pink background, impasto, color splashes, Cohesium,It's very basic. How would you prompt chaotic paint? I'm running around the circle here.
>>101249758I mean high frequency paint over this image. I guess I could try img2img.
>>101249758you might be somewhat limited by the model if you're using a finetune that over-prefers drawing coherent peopleas far as the prompt specifically, I've sometimes pushed token weights really high to get them sharply represented in a gen, like (abstract, surreal:2.0). you'll of course notice if you're pushing it too hard but exploring where the upper limit is can help get more juice out of the tokensanother approach is to add more texture/style tokens like "heavy brush strokes, lumpy paint texture, bold paint patterns" etc. not exactly related to your goal but lately I've liked to add (color gradients, double exposure) to lots of gens just because they produce some very interesting dynamics. if you're finding tokens have little effect, you can also try adding into pertubated attention guidance which can help align your gens closer to your prompts (though can also introduce artifacts if the scaling is out of whack)
I managed to make it closer to real pixel art by using python to make sure each pixel is uniform in size, what do you think?
How could I make videos like this?
>>101249945I think it all comes down to my own lack of English skill. And what I think what would represent the image. Thank you, I am a student here.
>>101249945Because I am the dreaded ESL.
>>101249955not trying to be a hater but the quantization is messy/lumpy/inconsistent. compare the blockiness of the eyes to the fine-grain of the outer edges, for example. theres also a lot of unsmooth jaggeds in a lot of areas>>101249967sometimes you can get a lot of success or ideas brainstorming with gpt. tell it what you want, ask it for improvements or better words, test them out in your imggen, and bring the shortcomings back to gpt to iterate. good way to shake out better approaches ot ideas you have
>>101250037overly smooth was what it was before, I set the minimum pixel size to 3 which is what causes that jaggedness, real pixel art often has that
>>101250037I think what also affects the result is the fact is the thing I more of a visual person, some people are just plain better in literal ways regardless of the actual language. English is an intermediate but I could have the same issue in Finnish language because I don't think about images in the same way.
I have always loved vanilla SDXL.
>>101245506Hello anons. I'm trying to gen using ComfyUI, but I get a lot of artifacts. I thought it could be the VAE, so I tested others, but I still see artifacts. How can I fix this?Also, to further contextualize, I'm using an artist's LoRA that works perfectly on tensor.ai but gives me artifacts on my local SD. Even if I remove this LoRA, I still get artifacts, albeit much less than with that artist's LoRA on.Here's a pic of how it comes out using that LoRA with my SD.
Luve me emiright is hassaku>>101250509Try add more steps, and use normal cfg range 5-8
>>101250550I decreased it to CFG 4 and it worked. I guess that's because I'm using AutismMix, though the gens still feel a bit trembly.
>>101250676>>101250685For these, this Koikatsu-made image (and corresponding socks one) was used in img2img AND canny, with denoising strength set to 0.7 and canny strength set to 0.4.I'm pretty sure I would not have been able to gen this without the Koikatsu reference - latent couple (and similar plugins) works just too inconsistently to get Haruhi, Mikuru, their corresponding outfits and especially the bondage correct.I first tried just img2img and with too low of a denoising strength it looked too 3D and with too high of a denoising strength it just fucked up... everything.Then I tried just canny and obviously it was a nightmare to get the colors/outfits all correct.Then I had the idea to load canny in img2img and it worked exactly as I wanted it to.tl;dr: loading canny in img2img can effectively 2D-ify a Koikatsu image
>>101250718I don't typically gen artifacts like that in Forge with CFG set to 7, so I think there's something other than just your CFG causing the artifacts.Are you stacking multiple LORAs, like an artist style LORA and a character LORA?
>>101247800SD3 is DOA because Civitai won't host SD3-based checkpoints/LORAs due to the license.
>>101250718>though the gens still feel a bit trembly.If the lineart seems not straight that's the indication of too low steps so you need to increase it. Except if you're using specified scheduler for low steps like LCM or PCM
>>101250762>Are you stacking multiple LORAs, like an artist style LORA and a character LORA?Yep. Does stacking them cause problems in local SD?
>>101249709Neat - it didn't gen any fog where it should be transparent.I've had mixed results with layer diffusion - it often (usually, actually) generates fog where it should be transparent for me.
This model probably gender biased. Afro/Cornrows will get you nigs. Mohawk will give you brown cunt>going on a trek to see gorillas in the wild
>>101250870Depends on the LORAs, but often, yes. Using multiple LORAs often deep-fries images.If you can use prompts to coax styles or characters out of your checkpoint instead of using LORAs, do so. Keep the number of LORAs you use at one time to the lowest amount possible.For Pony and Pony-based checkpoints like Autismmix, here's a spreadsheet of obfuscated style prompts:https://lite.framacalc.org/4ttgzvd0rx-a6jfAnd here's some style prompts that are recognized by name:https://civitai.com/images/6711496Note that some of those need underscores instead of spaces to work with Autismmix. (by aaaa_(quad-a):1) is one of those, for example. Doesn't work with a space instead of an underscore.For characters, just try using their character name as tagged on Danbooru and see if it recognizes it before resorting to a character LORA. For example, Autismmix recognizes hayase_nagatoro although it wants to draw her white so you have to add a dark_skin prompt, kek.
>>101250932>>101250870Also, I really like mixing style prompts. Try using two style prompts simultaneously with their strength set to 0.8 or three with their strength set to 0.7. You can create some unique styles this way.
>>101250932>aaaa_(quad-a)Shouldn't it be aaaa_\(quad-a\) ?
>>101250867I was using 12 steps, increased it to 20 steps, CFG 4. Still looks trembly, pic rel. I tried doubling the CFG and maintaining the 20 steps, but then the artifacts came back.Also, I'm using the option for normal scheduler. I'll poke around with PCM and LCM and see what I can get.>>101250932Tyvm, I'll see if I can gen better with your advice.
>>101251081Why would it be? Is this some weird comfy-only thing I'm not aware of because I don't use that autismal UI?
>>101251094Honestly 25 steps the minimum I'll ever use.
>>101251111To my understanding (quad-a) would be treated as separated word because weight bricket () is specified, not as one tag word. This is A1111 interpretation, Comfy even more confusing I have no idea about it at all lol.
>>101251145Alls I know that in Forge (by aaaa_(quad-a):1) works with Autismmix.
>>101251425Is this the rrealist?
Why is there a local diffusion and a stable diffusion thread?
>>101251437There's a lore behind it although I don't want to get into it
HiI am a poorfag thirdworlder, but my computer is really good, can someone here please spoonfeed me on how to do GOOD ai images? I used to hate them, but now the ones posted on /trash/ are so high quality that I can actually masturbate to them, I am not going to lie, I am a degenerate cuckold, I am going to use it to generate my favorite niche anime characters getting fucked by black men, so if anyone here would be willing to help poor me, I would be very happy as my life is kind of miserable, thank you for reading my post.
>>101251471Anon, do not go down the rabbit hole of using this shit for porn. I promise you it will be destructive like nothing else.
>>101251450Spoonfed me
>>101251490I can't jack off more than once a day because I have 0 stamina as a 14bmi loser, it's fine anon
>>101251471ask a catbox of the pics you like or search for catboxes already posted and load them in whatever interface you are usinghope things get better for you buddy
>>101251512I don't even know what an interface is, im asking mostly if there's a way to use novelai for free, honestly im fully clueless i just avoided all AI stuff for months until I started seeing really good stuff posted on /trash/
>>101251537as far as i know the two big ones are Automatic1111: https://github.com/automatic1111/stable-diffusion-webuiand ComfyUI: https://github.com/comfyanonymous/ComfyUIautomatic is easier to use, i don't know anything about novelai i never bothered with it
>>101251566Thank you anon, I'll look into it to the best of my abilities
>>101251471>I am going to use it to generate my favorite niche anime charactersSure, I'll hel...>getting fucked by black menNope.
>>101251437some people were like "sdg is cool but what if it took three days to clear a thread" so they made ldg
>>101251579I'm sorry for being white
>>101251568I'm not the most knowledgeable about these, but ask away if you need anything and I'll help you if i can
>>101251585Go see a therapist anon
>>101251471Download ComfyUI. It's super easy to use and super intuitive. You'll probably be able to quickly figure it out yourself and won't need a tutorial or to ask any questions or anything like that.As for what model to use, Stable Diffusion 3 is the latest and greatest model, so use that. It generates very high quality humans, both realistic and anime style, so it will be your model of choice for porn.Have fun, anon!
>>101251631Thank you anon, im currently installing automatic1111 but if my dumb little idiot brain isn't able to figure it out (i know nothing of programming) I'll try comfyUI
>>101251622Really nice!
>>101251646don't be intimidated. it just takes some time and patience and you'll pick it up
>>101251646this guy >>101251631 is messing you, comfy isn't easy to usehis whole post is full of shit
>>101251648thanks, you too
>>101251646this guy >>101251714 is messing youa1111 isn't easy to usehis whole post is full of shit
>>101251733This.
>>101251733Seconding this.>>101251714It's not very nice to reply to newbies with misinformation. Shame on you.
you guys need to make sense to me
>>101251733>>101251746>>101251754full of shit, you just want to mess with the new guyi've been using comfy for months now and i still haven't figured it out
>>101250718Post your setup to catbox and I can take a look.
>>101250718Problem is the fact any ponyXL makes shitty images. You should upscale them with sd15.
>>101251802>Problem is the fact any ponyXL makes shitty images.
>>101251820I did not ask your jpg or your opinion.
>hurr durr I didn't ask for your opinionYou're getting it whether you want it or not.
>>101251824I didn't ask if you asked.
>>101251846
>>101251810is this a reference to the science paper mouse with giant testicles? I laughed so hard at that shit
>Discord
I am angry and you are just clueless.
>>101251820What might be a skill issue for you might be a problem for people who actually get paid.
>>101251883she looks like a goat
>>101251904Ran recognized his animal fetishes. I do not think Ran has never had any paid work as an illustrator.
I am better than you.
>>101251863Yes
>>101251896
>>101251975been forever since we've seen this character
o
Good evening, anons! I hope everyone is doing well :]
>>101252095Ohayoo, PW-san!
>>101252095I was just posting my apple images.
>>101252095I was a week old image - the purple color has nothing to do with you. I swear to god. I welcome you and your energy is great in this thread.
>>101247916Working on my Blaze the Cat. Hopefully will update you all tomorrow about it.
>>101252095What's the most difficult dish you can make?
>>101252138How can you be this passive aggressive?
>>101252095hello pw, great to see you. pretty nice cup grip.
>>101252154ok Ill watch your 5 minute food video
>>101252184Marco is a legend my friend.
>>101252099Heyyy anon!! It's great to see you :]I actually tried to get on in the morning but it was WAY too hot hahaha!>>101252109Nice, anon! It looks just like him!>>101252123I think it looks great haha! Aww thanks so much, anon! :'D>>101252138Uhmmm I never thought about that LOLI can do quite a few a few different things tho :]>>101252150Heyyy Debo!! It's so great to see you again! I hope you're having an amazing day so far!
>>101252203How are you thinking about this Mr. White video? What is your touch?
>>101252191>doesn't show the finished dishREEEEEEE>>101252203are you working on the 4th?
>>101252211Mr White video? Whacha mean haha>>101252224Nope!! I have it off :DI should be home too haha but we'll see! I didn't get any fireworks but my friends did haha
ok, both finished downloading at the same time, comfyui's ui does seem a lot simpler, but i dont know how anything works, i tried copying and pasting some random porn from rule34 and i got like 2 clay dolls, where do i feed it images to steal from?
>>101252293Thank you for listening.
>>101252297oh jesus i tried the other thing and used img2img and it ended up like this
>>101252224The dish is like an any artists interpretation.
>>101252325its perfect
>>101252131>>101252395Lovely>>101252203How do you like your steak?
Hows SDG in /g/ going these days? I usually hang around in /e/. You guys got any weird schizo posters over here. We have a couple unfortunately
>>101252421>You guys got any weird schizo posters over here.nope, just 100% normal non-trans anons here.
>>101252421we probably share the same schizos
>>101252421What is the problem here then?
>>101252448saar?
>>101251589I would like your help
>>101252452Your mind is so filled up with trannies and Indians I am genuinely interested how do you do in the US?
>>101252456what do you need buddy?
Is Forge still supposed to be way faster than auto1111 (dev branch)? I'd heard some stuff about performance improvements having been ported over, but on Forge with SDXL I still seem to get 2x performance over auto1111 master and about 1.5x performance over dev branch.Just unsure if I've got some fucked settings/environment or if this is more or less expected for now.
>>101252463uhhm.... saar?
>>101252480Just choke off, kill yourself.
>>101252474I don't really get how to feed it images, like how to train it, the results are all just scrambled messes, can you explain how i can do it? i can do it slowly and manually adding 1 image each by eachcomfyui or auto is fine, i have both downloaded, whichever one you choose to explain i'll use
>>101252485Nta, but rude.
>>101252421we're on 4chan so of course we've got schizos. a bunch of the worst ones left recently though
>>101252485fack you, you bitch mother
>>101252497Of course it is rude.You are rude just by existing with your simple minded idea about English and US. Please burn in hell.
>>101252505what did pajeet mean by this?
Is running ComfyUI on vast.ai the cheapest option for SD if I don't have a high VRAM GPU? It's worked okayish so far but kind of buggy. Also any tips on making dickgirl hentai?
>>101252501At least I learned >Swedish>German>EnglishWithout my family going bankrupt.What is YOUR second language?
>>101252514>Also any tips on making dickgirl hentai?Paging Dr. AnimAnon. You're needed in the futa ward immediately.
>>101252513What is your second language, Sir?Or is this the psycho avoiding ip bans?
>>101252516what language do you prompt in?
>>101252533Hindi, my friend.
>>101252488I've never trained a model or a Lora, and i don't think that's what you are looking for either. Just download a model and loras you like from here>Models, LoRAs & upscalinghttps://civitai.comhttps://huggingface.cohttps://openmodeldb.infoand start generating pictures
>>101252524>>101252516>unironically seethingif the curry is too hot, get out of the bazaar
>>101252410Thank. You are Cardosface? Good to see you. Is goth anon.
>>101252556that's troony
>>101252551Oh here is someone who thinks English language is somewhat mystical, and someone who has never traveled outside his particular state in his life. You know what other people call you? You are a hick.
>>101252551Fantasizing about Indians, you have never had any single meal in London in a real Indian restaurant.
>>101252564>You know what other people call you?saar, usuallylmao got 'em
>>101252563Yeah, but I think he used to be cardosface before that, they are the same.
STOP IGNORING me is vast.ai the best for poorfags? It seems pretty good overall. And HOW do I make my DICKGIRL HENTAI?
>>101252571You are the schizo again. Please try to capitalize your phrases please.
>>101252488>>101252543also you can download some example pictures from the model you like in civitai for example and drop the image in this tab and then choose "Send to txt2img" and you can make any changes if you want to the prompt and the all other the settings
>>101252571I told you before, you are too stupid to vary your own English.
>>101252570i prefer when your mother cooks it for me>>101252579>>101252590keep seething, saar
>>101252612If you ever travel to London (I doubt you do because you are a hick)https://www.google.com/maps/place/Punjab+Indian+Restaurant/data=!4m7!3m6!1s0x48761b32d462063b:0xf4cf9b277316dff4!8m2!3d51.5151154!4d-0.1265625!16s%2Fg%2F1tj5kmx2!19sChIJOwZi1DIbdkgR9N8Wcyebz_Q?authuser=0&hl=en&rclk=1Take all the trackers with you.
>>101252564nta but I've been to the majority of states, mexico and [french] canada. have I graduated from hick?>>101252578none of us pay for cloud services :(>>101252579nta but capital letters are a government psyop. every capital letter you skip, the less power the government has over you. also, never prompt with capital letters. they lower the quality of your gens
>>101252636You only know if you have been graduated from the hick. It is a spiritual experience my friend.
>>101252629only a stinky pajeet would mald this hardi look forward to licking the curry off of your mother's fat ass, oh Shiva halp me
>>101252612Fuck you hick.
>>101252543>>101252585still did not work... can you spoonfeed me more? i added the lora to sd.webui\webui\models\Lora
>>101252660Fuck you hick.
>>101252660Oh your cracker father did not teach you how to type? It is not my fault your father beat you and the result is a psychopath.
>>101252636>none of us pay for cloud services :(because you just have a good gpu?
>>101252674>>101252681you smell like the Ganges, saarplease bathe in clean water
>>101252662can you upload the picture to https://catbox.moe and post the link here?
>>101252711I am from Finland. You have never travelled in your life.
>>101252711Maybe do a little psychopath page on tiktok or something?
>>101252718>I am from FinlandI admire your country's autism, I've always been a big fan
test
>>101252729oh good one saar, very good>>101252718>I am from Finland.ah, I would be mad too
>>101252738welcome back
>>101252712the pnginfo thing you suggested didnt work so im just trying to do img2img again with afrobull images (yes yes boring i know its just the only artist im aware has ai models), do you think its a problem with the prompt? im using this website someone linked me on /trash/ but most generated images i find on the internet seem to have the metadata wiped https://novelai.net/inspect
>>101252578No, it's one of the worst cloud offerings because they charge for egress AND ingress, aka it costs more to download your models and upload images out of the instance than the hourly rate
>>101252768yeah it's been pretty annoying and it constantly runs out of space. suggestions on something better to use friend?
>>101252740>>>>101252718(You)I just farted and it was better than anything you have ever done in your life.
>>101252783Runpod don't charge for bandwidth but have the similar issue of charging too much for storage. I'd use one of the free saas or just not bother, like if you're that poor you can't afford to buy a 3060 or anything then surely you've got more important things to spend your money on.
>>101252821i'm just a student and don't want to buy a pc right now but i find SD fun. i guess saas is probably the way to go yeah
>>101252752beep beep boop?
Everyone post the most impressive, in your opinion, Stable Diffusion image you've ever made.
>>101252851blue board
>>101252821also wouldn't a 3060 be a huge piece of shit irt ai image generation? i thought you need at minimum 24gb vram
>>101252860Send it to me on discord wingclipper0
>>101252851I've made too many impressive gens. no idea what the MOST impressive would be.
>>101252870no
>>101252879bizarro debo was a good series too
>>101252712>>101252758ok i tried it again, i looked some images up on a booru and got it to work img2txt, so i edited some stuff around and got this horrible incomprehensible image: https://litter.catbox.moe/g1i9by.pngwhat did i do wrong? if it helps, as you can see, i also downloaded a lora for the character i wanted to generate but it didnt show up here, does that have to do with anything?im sorry for being stupid and retarded and asking to be spoonfed so much, im just bad at computers, also my electricity stabilizer keeps clicking when i generate something, is that normal?
>>101252896>>101252896>>101252896
>>101253004
>>101252944try using a smaller resolution 640x928 for example and then upscale to 1280x1856 by using the Hires.fix and see how that goes
>>101253017
>>101253030
>>101253039
>>101253030prompt/workflow?
>>101253018answered you here since the thread image limit is full >>101253078 sorry
>>101253079its not as cool as you might hope. it pretty much produced the same picture over and over again. I couldn't get much variability/creativity out of it, but maybe you canhttps://files.catbox.moe/ytmlcw.png