Previous /sdg/ thread : >>101970411>Beginner UI local installEasyDiffusion: https://easydiffusion.github.ioFooocus: https://github.com/lllyasviel/fooocusMetastable: https://metastable.studio>Local installAutomatic1111: https://github.com/automatic1111/stable-diffusion-webuiComfyUI: https://github.com/comfyanonymous/ComfyUIForge: https://github.com/lllyasviel/stable-diffusion-webui-forgeSD.Next: https://github.com/vladmandic/automaticAMD GPU: https://rentry.org/sdg-link#amd-gpuIntel GPU: https://rentry.org/sdg-link#intel-gpu>Use a VAE if your images look washed outhttps://rentry.org/sdvae>Run cloud hosted instancehttps://rentry.org/sdg-link#run-cloud-hosted-instance>Try online without registrationflux-dev: https://huggingface.co/spaces/black-forest-labs/FLUX.1-devsd3: https://huggingface.co/spaces/stabilityai/stable-diffusion-3-mediumtxt2img: https://www.mage.spaceimg2img: https://huggingface.co/spaces/huggingface/diffuse-the-rest>Models, LoRAs & upscalinghttps://civitai.comhttps://huggingface.cohttps://aitracker.arthttps://openmodeldb.info>Black Forest Labs: Fluxhttps://huggingface.co/black-forest-labs/FLUX.1-schnellhttps://comfyanonymous.github.io/ComfyUI_examples/flux>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>Related boards>>>/h/hdg>>>/e/edg>>>/d/ddg>>>/b/degen>>>/vt/vtai>>>/aco/sdg>>>/trash/sdg
>mfw Resource news08/20/2024>GGUF conversion of Google's T5 v1.1 XXL encoder modelhttps://huggingface.co/city96/t5-v1_1-xxl-encoder-gguf>hydit-live: Minimalist Hunyuan-DiT Gradio UIhttps://github.com/geoffitect/hydit-live/>Anim-Director: A Large Multimodal Model Powered Agent for Controllable Animation Video Generationhttps://github.com/HITsz-TMG/Anim-Director>TraDiffusion: Trajectory-Based Training-Free Image Generationhttps://github.com/och-mac/TraDiffusion>Procreate’s anti-AI pledge attracts praise from digital creativeshttps://www.theverge.com/2024/8/19/24223473/procreate-anti-generative-ai-pledge-digital-illustration-creatives>Comfyui_TTP_Toolsethttps://github.com/TTPlanetPig/Comfyui_TTP_Toolset>SimpleTuner Adds Lycoris Support for Fluxhttps://github.com/bghira/SimpleTuner/releases/tag/v0.9.8.308/19/2024>Safetensor UNet Loader / Merge Plugin for AUTOMATIC1111 Stable Diffusion Web UIhttps://github.com/captainzero93/load-extracted-unet-automatic1111>Generative Dataset Distillation Based on Diffusion Model https://github.com/Guang000/BANKO>Efficient Image-to-Label Diffusion Classifier for Adversarial Robustnesshttps://github.com/hfmei/IDC08/18/2024>AI + a16z: The Researcher to Founder Journey ft. Black Forest Labs foundershttps://podcasts.apple.com/au/podcast/the-researcher-to-founder-journey-and-the/id1740178076?i=1000665678592>FLUX.1-schnell-training-adapter: Train LoRAs directly on schnellhttps://huggingface.co/ostris/FLUX.1-schnell-training-adapter>XLabs-AI / flux-controlnet-canny-v3https://huggingface.co/XLabs-AI/flux-controlnet-canny-v308/17/2024>FastSD CPU v1.0.0-beta.36 release with FLUX.1-schnell OpenVINO supporthttps://github.com/rupeshs/fastsdcpu/releases/tag/v1.0.0-beta.36>UNet Extractor and Remover for Stable Diffusion 1.5, SDXL, and FLUXhttps://github.com/captainzero93/extract-unet-safetensor
>mfw Research news08/20/2024>SANER: Annotation-free Societal Attribute Neutralizer for Debiasing CLIPhttps://arxiv.org/abs/2408.10202>MeshFormer: High-Quality Mesh Generation with 3D-Guided Reconstruction Modelhttps://meshformer3d.github.io/>Fairness Under Cover: Evaluating the Impact of Occlusions on Demographic Bias in Facial Recognitionhttps://arxiv.org/abs/2408.10175>Video Object Segmentation via SAM 2: The 4th Solution for LSVOS Challenge VOS Trackhttps://arxiv.org/abs/2408.10125>Factorized-Dreamer: Training A High-Quality Video Generator with Limited and Low-Quality Datahttps://arxiv.org/abs/2408.10119>CLIPCleaner: Cleaning Noisy Labels with CLIPhttps://arxiv.org/abs/2408.10012>Caption-Driven Explorations: Aligning Image and Text Embeddings through Human-Inspired Foveated Visionhttps://arxiv.org/abs/2408.09948>Attribution Analysis Meets Model Editing: Advancing Knowledge Correction in VLMs with VisEdithttps://arxiv.org/abs/2408.09916>RealCustom++: Representing Images as Real-Word for Real-Time Customizationhttps://arxiv.org/abs/2408.09744>MePT: Multi-Representation Guided Prompt Tuning for VLMshttps://arxiv.org/abs/2408.09706>Photorealistic Object Insertion with Diffusion-Guided Inverse Renderinghttps://research.nvidia.com/labs/toronto-ai/DiPIR/>StyleBrush: Style Extraction and Transfer from a Single Imagehttps://arxiv.org/abs/2408.09496>Classifier-Free Guidance is a Predictor-Correctorhttps://arxiv.org/abs/2408.09000>RepControlNet: ControlNet Reparameterizationhttps://arxiv.org/abs/2408.09240>Flatten: Video Action Recognition is an Image Classification taskhttps://arxiv.org/abs/2408.09220>Are CLIP features all you need for Universal Synthetic Image Origin Attribution?https://arxiv.org/abs/2408.09153>StylePrompter: Enhancing Domain Generalization with Test-Time Style Priorshttps://arxiv.org/abs/2408.09138>Thin-Plate Spline-based Interpolation for Animation Line Inbetweeninghttps://arxiv.org/abs/2408.09131
First for Debo is the thread schizo
Is Flux good? can you use it with Automatic1111 webui?
>>101987347Yes, it's good. No, but you can use it with Forge, they have the same interface, and if you install Forge you can link all your a1111 models and loras to it.
>>101987365Oh wow, I've been out of the loop really. What's the best one to use nowadays? A1111, ComfyUI, or Forge?
>>101987390After the last update, I like Forge currently. A lot of people like Comfy, just too many missing custom nodes for me and always run into errors, can't just copy someone's workflow easily. You definitely should try Forge, I think it's probably as good as Dall-e 3.
>>101987433I meant try Flux, with either Forge or Comfy.
Wow, I can't get OneTrainer to output anything usable. Steps apparently aren't a thing anywhere (or "epochs" are steps now, I don't know) and it's really fucking flaky with various setting interactions. Kohya was old, esoteric garbage, but at least it worked.Kinda like going from A1111 to Comfy; everything is reorganized strangely and super touchy (wanna fix a red box? reinstall) to the point of annoyance for no reason other than to do it differently. lol
>>101987604I installed it and never used it. I just used the kohya fork, but a lot of times my loras were crap. Don't think I ever got the settings really right.
Anyone using Stabillity Matrix? Is there more performance if I use the usual local instation for each package?I've recently moved my loras and models, reinstalled packages to a new partition my models(on a portable SSD, cRUCIAL X10 PRO) and it starts so slowly compared to before! Gives the impression it's stuck on this screen
What's the best value vid card for Flux? My 8gb 1070 barely runs Pony on Comfy, so anything is an upgrade at this point. Flux checkpoints are huge, so 16 gb is min?
>>101987654I never knew this existed
Whats the difference between /sdg/ and /ldg/?
Good evening, anons! I hope everyone is doing well :]
>>101987769It's just different cliques from discord I think, I never joined the server so I don't know the details
>>101987769Some anon wanted to change the name to be more inclusive of other models. Some people didn't want to and wanted to keep the name so they stayed here.
>>101987808I guess this too >>101987806 The people in /ldg/ all hang out on discord together.
>niggers tongue my anus :]
>>101987788Hello
>>101987765>>101987845These are really nice, in b4 shikanokokokodo>>101987788Cursed backrooms PDub gensAlso I have nothing to posts so have uuuuuh this thing
>>101987845Hello, anon! Cute gen :]That's a really nice styleI hella like her ears and antlers haha>>101987864Mouse anonnnnnn!! Hello :DLOL same actually, i'm doing a LoRA for a friend right now so i'm kinda just using stuff I already made hahaI like the detail of that crystally thing! It looks really nice :D
>>101987864Thanks, nice crystal tree>>101987905I didn't notice she had 4 ears in that one
>>101987944I think it's cute haha!Her hair style is really cute too :]>>101987986Nice teddy bear, anon! That's a cool style!
>>101987390comfy is the same thing as auto and forge and just implements some bad node interface, if only chainner had better SD support.
>>101988035ty friendwhat is your witch up to, why is she playing alone in a lan?
web scraping 102 droppedhttps://huggingface.co/blog/hlky/web-scraping-102
>>101988063Any time!! :]Alone for now! This arcade is abandoned hahaha>>101988169Hlky! I hope you slept well earlier hahaWelcome back :]
>>101988035Thanks!>>101988169Cool, I'm going to give this a try some time.
LORA Training question: I have a large dataset full of very dynamic images, but I realized too late that they all have the exact same background. How do I best mitigate the background without manually cutting the subject out of every one? Can I like, train on the background alone and then negative-merge it? Is that even a realistic proposal?
>>101988202On 1.5 that will probably be a problem. On Pony it doesn't seem to be. Tag the background and everything in it as much as possible. If you can add just a couple with the bg cut out it should help.
Flux is kind of good at recreating that shitty pixel lora from like 6 months ago
flickr 482m! will drop an update at 500m>>101988185yeah i slept reasonably well. hope you had a good time at work and have had a good evening!>>101988187if you do i'd be interested to know what you think. writing these educational articles is something new i'm trying
>>101988227That's heartening, thank you. About one in four of the dataset are other images so I'll just do an extra thorough job on the tags.
>>101988251Okay, I'll let you know. I'll try it next time I make a dataset.>>1019882591.5 would get stuck on a repeated concept. It could also learn one easily. I made this Daria lora and there were only 2 images of her from Beavis and Butthead and it learned that style. Pony can't do that. But yeah, good tagging I think is key.
>>101988357Without the Beavis and Butthead tag
>>101988357>>101988374Wtf Daria
>>101988251I'm glad to hear that! :]I did! It was a long day haha but it was good! Thanks so much! I'm having a great time hahaNo work tomorrow too so i'm super happy about that!
>>101988381Yeah, somebody asked for it. I had a bunch of other retro cartoon waifus.
turns out /ldg/ has some resident schizos so im back home
>>101988435I really haven't browsed it in a while
>>101988429>all that ever matters in a relationship is that you like the same pizza toppingsDaria had the best one-liners
Are there any kind anons that can please explain why Flux based lora's dont work for me? Others seem to work fine.For example this one does not influence my outputs at all: https://civitai.com/models/649980/cleavagehelper-flux?modelVersionId=742855So for example adding the following LORA to my prompt works: <lora:breastsizeslideroffset:0.8>, <lora:Bdelphine760x1024v2:0.6>, But when I add the following flux based it has no impact: <lora:CleavageHelper-FLUX-rc4:5I'm using automatic1111 with the checkpoint flux1-dev.safetensors Any advice frens?
>>101988357it's more for education desu, still good to get the latest data but if you just want data from civit there are datasets availablehttps://huggingface.co/datasets?search=civitthese two are the most up to date, mine is the first one and includes the actual images in webdataset, i should upload the metadata separately because hf's dataset viewer has an issue with this set, then you could filter it, but you can also just grab a few shards and see what's in themhttps://huggingface.co/datasets/bigdata-pw/civitaihttps://huggingface.co/datasets/AdamCodd/Civitai-8m-promptsi'll be covering efficient bulk downloads of urls from parquet datasets in a future articleand if you want existing training datasets uploaded to or trained on civit, check outhttps://huggingface.co/datasets/bigdata-pw/Dataceptionthese are in webdataset which i am planning on adding support for to kohya etc. but you can also simply extract the tar files in the mean time>>101988405that's great, enjoy your time off!is that image from your lora? it's looking good
>>101988435furniture anon was fun
>>101988545It is something I want to learn more about. You could probably put it on civitai also since they have articles.
>>101988545Will do! :DYeah it is, thanks so much!!>>101988435Welcome back, anon! :]
>>101988673>not barefootyikes my dude
>>101988617>civit article about scraping civit with the non-public apii don't think they'd be too happy about that
>>101988435Which one? The CP poster made me nope out of there for a while
>>101988673>>101988692Pretty good fire horses>>101988703Yeah, that's true. I was just thinking people might find the scraping article helpful, but probably not with that part.
>>101988712the ones fighting each other cluttering up the thread with text to the point theres barely any gens
>>101988528Yes. I didn't realize until recently Mike Judge wasn't that involved with it.
>>101988202Add noise to all images.
/sdg/ deserves some flupuszuch to
>Maintain thread qualityhttps://rentry.org/debo
Finally got the shirt (sort of) right
>>101988820what is it?
>37 downloads on flickrnicebigstockimage33m webdatasets are almost ready, it's ~1.4tb though so the upload will take at least a few hours>>101988735the principles learned in the articles apply to almost all web scraping so hopefully people will find it useful. i'll cover some more real world examples in later articles
>>101988778Funny you should mention. I was just discussing Office Space and B&B the other day with my boss because we have a coworker that looks exactly like Milton and I always laugh forever when I see him.
>>101988930prompt so I can use in different style and model?
>>101988895a flumpuszuch with its babies
>>101988990>masterpiece, highres, 1girl, solo, short black hair, closed eyes, smile, purple robe and beret, tired, sitting in front of computer, drinking coffee, hand on face:]
+zoom
>open up my a1111 webui >load up my pony diffusion model >Disney princess lora at full strength without prompting a disney princess? Check. >Add detail at 1.5?Check >3 different negative embeddings,one of them for SD1.5?Check.>masterpiece:1.3,ultra detailed:1.1,godrays,perfect, beautiful, sexy girl?Check?>bad awful mutated hands wrong fingers fucked up anatomy?Check. Yep, it's proompting time.
>>101989109>>101989117-zoomhttps://files.catbox.moe/4uknv0.png (nsfw)>>101989122you could edit your\stable-diffusion-webui\ui-config.jsonso your default options are set, and you just have to click gen q_q kek
flux ctrlnet works, but god damn its slow
>>101988777they are the one and the same, and i also don't blame anons for not recognizing looping llm slop.
>>101989024they are cute, but what are they
>>101989109>>101989117>>101989143>>101989236I like
>>101989258just flumpuszuch
Last one from me, good night anons
>>101989353Good night, Quokkanon! Sleep well :]
>>101989340are they edible?
>>101989353Gn quokkers
>>101989097thanks for sharing buddy
>>101989509maybe
>>101989681Omg that looks amazing, anon! :DI love it!!
just smashed out a scraper for top games from steam then their screenshots in like 5min desu
>>101989977and now what? what's your next move going to be?
>>101989977Ohhh nice!! which games?
>>101990006after i've got a BIG dataset you're going to tag them with characters and other game specific features for me, thanks! your effort is appreciated>>101990021everything under topsellers https://store.steampowered.com/search/?filter=topsellers
>>101990021I hope the unreal engine 5 port they're doing ends up looking half as good as your gens.>>101987905love the sneaky skill icons in the background on this one.
>>101990055Sounds like you need some coffee.
>>101990179dw i have coffee>>101990187i'm looking for a good source of console game screenshots, if you have any ideas let me know!
>>101987505>>101987765>>101988169>>101988966Adorable
>>101990086>>101990093Very nice
Do you know what you call a gay soccer player who plays at professional level?A star playerIt works better in Dutch sry
>>101990187>This looks SO good, anon! :DGlad you liked them friend, one last purple witch from me.>>101990386Thank you
big load
>>101990187I've been playing since October 2010 and only took about 5 months off, still playing right now and our boys have a guild and thread over on >>>/vm/ if you're interested in playing with other anons when you come back.We have a couple others that play with SD in guild but I'm the most /g/ inclined of the group, mostly lurking over here since a little before the NAI leaks but you got me to post!
>>101990362I'll try to think of something! I haven't played much console stuff haha just PC>>101990399I really do!! Thank you!I'll save them forever :]Such a cool style!>>101990473Niceee!! I've been playing since beta til uhhh maybe like 8ish years ago hahaha! I got super fashionogi about it though LOL so i'm just okOhhh nice!! I would love to! Btw, what server did you start on before the merge? Your char is so cute! Here's mine :]Though I guess it's obvious since I made the LoRA LOOL
>>101990597I'm from Tarlach, most of our other anons are Alexina or Mari though. I use to main an elf but switched over to giant when they did the close combat buffs for them (but mostly do mage again now anyway) and have mained her since, glad you like her :)Also fashinogi is at least half the game! I might be 73.8k total but I do way more doll simulator and yapping with the boys than grinding endgame dungeons.
I updated stability matrix and now there is no DPM 2++ Karas, why?
>>101990875Oh cool!! I'm from Ruairi hahaWow nice!! Idk what lvl I am, but it's likely nowhere near that LOLI was mainly an archer but dabbled in other stuff as it came out hahaLOOL it really is! I spent so much time in dunby doing literally NOTHING on a daily basis hahaha!I would love to get into all the new content! I think I stopped playing around the romeo and juliet stuff Suuuuper long ago!
why is the other thread so much redditi feel dirty just looking at it
i miss schizo anon
>>101991150This is a tranny paradise
>>101991355all straight white men over 6 ft tall are here
Im so happy all the avatarfags pooped very well todayI was worried for a moment
>>101991409Me too, cheers
>>101991150I have both open and both have annoying parts, but I am here for interesting gens and tech info so I guess I have to be ignoring the annoying sides
>>101991409Well, constipation is a serious problem.
>>101991343>>101991507thinly veiled racism
added some contrast
>>101991409I actually had a great one today, Ty for worrying king! Hope yours goes well too!
Stanley steamer keeps your home cleaner
Stability AI Is dead, isn't it?
>>101992781As dead as /sdg/ and the romance between Julien and cumfart
>>101992821Who?
>>101993018more like whooo
Oh no! Someone got mad about my criticism about /ldg/ hahaha!!It's been a while, hasn't it? :](not an avatar btw)
>>101992571How does this compare to a Cleveland Steamer?
>>101993097Porn has destroyed your brain
>purple schizo banned yet again for avatarfagging>his solution is to break the rules yet again and ban evade>:]
>>101993079What did you get banned for this time?
>>101993117these >>101992571 >>101993097 are porn references?
>>101987505>>101987765>>101987845>>101987944>>101988187Aren't antlers only for male deer?
>>101993131>>101993149Hahahaha it was just a warning, as usual :]https://suno.com/song/4907bcaf-af7f-4415-ba86-8044a46856a3
>>101993149Avatarfagging and stirring drama like the last 100 times
>>101987806>>101987818what discord?
No surprise all good posters left this shithole
>>101991150Imagine using reddit as an adjective
>>101993177Let me guess; another warning for avatar/signature usage?
>>101993041whooooo
>>101993177Rulebreaking avatarnigger
>>101993291You know what to do
New wallpaper.
>>101993311Way ahead of you schizofren
>>101993219Awww you tried, anon! Gold star!(not an avatar btw)>>101993247You would know, anon! :]Better luck next time!(Totally not an avatar btw ;])>>101993291Do something about it, loser!(Not an avatar, of course!)https://suno.com/song/df44f804-0b0e-4da3-badc-392d88fea7a9
>>101993326hey that's my house
I think she operates on the assumption that the average anon is dumber than her but I don't believe that's true
>>101993387Its one of the only anons still talking to debo after all he has done so some kind of brain damage is implied
>>101993387I don't even think anyone reported the purple schizo
we're so back bois
wait so reporting for avatarfagging and trying to convert the general to a subreddit actually works? thats very good to know>:]
BIGstockimage35M droppedhttps://huggingface.co/datasets/bigdata-pw/BIGstockimage35M
>101993439gm Debo
>>101993465Just do it on sight and don't engage him
hes resetting his router one moment
>>101993510will do from now oni was under the impression that it doesnt do anything so unironically good to know>:]
>>101993467so many asians in that one lel
>>101993600you're unironically schizo anon aren't you? kek
>>101993651we are all schizo anon but theres only one thread schizo (debo)
>>101993601the preview only covers the first 117k rows, it's actually a very diverse set, the initial scrape used combinations of categories including ethnicity and just due to the ordering the first rows do have a number of indians. after the category based scrape it went through a large list of locations/countries
Let's try something new hahaha!!The imglets sure woke up early today!
>>101993732Is that the new name for nogens?
>>101993746Not new at all! I've been saying it for at least 2 years hahaha!Nogen sounds too nice for their kind!>https://suno.com/song/5dcf52ff-17c3-4f11-99db-59ebdf452827
I hope koff is ok. we haven't seen him in a while>>101993732good morning purple haired poster>>101993733so we know how they're turning the frogs gay, but where are they getting the gay juice in the first place?
>>101993824Thread schizo
>>101993794That makes sense.Also, they've been at it for 2 years?
>its another "gromming a newfag" episode
>>101993896who is the newfag who are the groomers?
>>101993824Good morning anon i've never seen before!! It's great to meet you :]>>101993844What a sad life you live :]>>101993866Yup! /sdg/ used to be super cool and everyone was friendly until they showed upSome think it's /ic/, but I think it's the resident /g/ losers, or maybe those who wanted /ldg/ but still bother usOh! I got a warning for bringing them up earlier tho! Good thing i'm not using an avatar tho haha!
>>101993915The ones who are asking questions and the ones who are crying with a trolljak mask on
>>101994015Nice old pc games effect
He's having a rough one
>>101994015very pretty, want to share prompt?
This should be enough, see you all in 10 mins or so :]
>>101994167take care buddy
>>101994038thanksI love good pixelart, but SD/Flux is ass at generating directly pixelart, and in all honesty it shouldn't even try, we got specialized networks for pixelart, and they do a much better job than what SD/Flux can generate>>101994092it's a simple prompt that I gave it to llama-3 to improve it, and I just slapped some wildcard artists on tophttps://litter.catbox.moe/wdj0xi.png
>>101994167see ya
>>101994167>10 mins or so :]Take it easy dude!
KIX is the best cereal
>>101994297Fruit loops for the win.
>>101994336Objectively incorrect
>>101994297berry kix or normal? I got this new strawberry protein powder and was thinking "this tastes familiar" and realized it tasted just like berry kixI like cinnamon toast crunch but haven't had cereal in a billion years. I kinda wanna gen cereals now but idk how to make it interestingalso, who is this lady you keep prompting?
>>101994422Normal KIX. And its just a woman.
Morning anons
>>101994624gm
What a silly image: https://files.catbox.moe/l0wdbk.png>>101994200Thanks so much! I'm back now!>>101994293Thanks!!! Cool gen!>>101994295Omg is it you? If so, I love you hahaha!>>101994624Good morning, Quokkanon! I hope you've slept well :]
>>101994657>Omg is it you? If so, I love you hahaha!Yes it's me! I thought you typo'd 10mins and meant hours!
>>101994657>What a silly imagethats an avatar? thats pretty silly. you could add a "wearing __color" wildcard and be immune xDits also pretty funny how bitter ldg is about the truth. have they ever filled a thread?
>>101994685LOOOLYou're the greatest, anon!I'm gonna be here for a while, hopefully, then I'm gonna run for a bit hahaha!I appreciate you!>>101994750I guess LoRAs are "avatars" now? Who knew!!! LOLikrrrrThey never have hahaha! But shhh they get real mad about that! Cause it's true!
Was my jam>discontinuedDamn son>>101994624Goo mornin'
2m video game screenshots so far!493m flickr, 500m soon:tm:also rate my landing page, name hidden until it's ready for launch. i'm building the api and other stuffsoon:tm:trust the plan
>>101994850Mouse anonnn! Omg I had cereal for the first time again in years a couple days ago!I had honey bunches of oats hahaha!Waffle crisp is hella good too!>>101994914That looks super cool, Hlky!!!
How to detail only the largest face with facedetailer on comfyui? This shit is so complicated.
has ani even posted in the past few days?
>>101994945ani is locked up in comfy's basement
>>101994930I just eat boring keto cereal and almond milk now
>>101994655Morning :)>>101994657Hello PW, i did sleep well, better than yesterday now that i managed to get flux working locally.
>>101994850Hello, nice breakfast :)
Heck yes, Simon Stalenhag! FLUX understands it naturally, you need no lora.
>>101994930thanks!! the whole thing is going to be super cool. if i drink enough coffee i think i can have it ready in two more weeks or less
>>101994982Aww but at least other keto foods are good tho!You can still have bacon and eggs for breakfast!My cold brew uses almond milk haha it's pretty good!>>101995049I'm so glad to hear that, Quokkanon!!Ohhh nice! That's a very cute quokka :]Flux is super fun to use!>>101995089I can't wait to see it, i'm actually super excited for it!!I need to get you a coffee hahaha!
oldiein the psych ward, last few daysdoctor says he thinks i am on the autistic spectrum... shocking..had a dream that consisted of fucking around with samplers
>>101995202Fo sho'>>101994914Bretty cool>>101995369I had a dream last night a cute tomboyish girl busted through the wall and took me up through the Backrooms where she lived for a date but the floors kept glitching out and disappearing
>>101994914>also rate my landing pageeach character could have a rand interval where they shift into a different character. they could also have a slight opacity proportional to scale so as to create a sense of depth>>101995082I hope someone puts together a list of artists flux knows. its a crapshoot rn. its actually extra weird because there's some artists it seems to know but then it also doesn't produce their style. like, the artist token changes the style towards that direction but not actually in their style... if that makes sense>>101994945he doesn't hang out here anymore. too cool or too busy for us, I suppose
>bmp ignores nigbo again
>>101995408despite my obsessions, rarely, very rarely, have 1girl dreams. mostly wandering buildings.
>>101995202i like it milky and sweet>>101995369i have lots of weird dreams. a recurring one recently is being in a huge futuristic train station, i can't find the platform i need and i'm worried i'll be late home for dinner>>101995408ty!>>101995425good idea. i'll add it to the backlog!
gonna 1girl so hard when i get out
>>101995535Me too!! Hella cream and sugar hahaha!
>>101995544What are you in for?
mfw ageslider WILL be at -3>>101995584the usual, suicidal thoughts
>>101995535>add it to the backlogwhere ideas go to die D:>>101995544glad you're not dead. we get nervous when you disappear
>>101995615>mfw ageslider WILL be at -3
mfw and weightslider WILL be at -2 or -3, and the images WILL be uncanny
>lost city of Atlantis, psychedelic >>101995472Exploring dreams are fun. For me it's always up and never down for some reason.
Trying CuteSexyRobutts style, and while it looks half-decent when theres no limb abomination, it looks a bit washed up...Any tips on how to correct this? I am artfag as well, so I will try to do something in CSP, but wondering how someone with more expertise in AI would approach this illustration to improve it even more.
>>101995839Vae issue maybe. If all else fails just throw it into gimp and turn up the saturation
>>101995924Thank you, gonna check this out. I thought VAE wasn't that important since everyone seems to have it baked, but I will check the configs.I supposed the wobbly eyes is something I would need to fix in CSP alongside the artifacts and other stuff.
>>101995581i used to have 4 sugars, now i add a drop of honey, it's much nicer>>101995646dw ill do it before launch. there's just so much to do!
I was trying to get realism on the book since I saw someone in /ldg/ get anime face on realistic body but I couldn't get it to work>>101995963np
>>101996130Mean robot, not body
>>101996130You sure they ain't using ControlNet? Usually is the only way to get gimmick stuff like this.
>>101996051Why not just plain black? I think you might be addicted to sugar... We need to cut down your coffee consumption.
>>101996301Not sure, it was >>101989273
we have 15minutes to make it to the next thread or I'll miss the news
>>101996319black coffee is disgusting
>>101996532that just means you've never had good coffee before
I give up
>>101996769Nice. She's trying to hide some alien features underneath those sunglasses...
>>101996769most of them look very real
>>101995581>>101995535you people disgust me...adding milk and sugar to already diluted espresso...mamma miaPorca miseria!
>>101996932I put ketchup on my pasta
>>101997016Oh I didn't know he's a real person. I was referring to LA Confidential.
>>101997002I would disown you if you where my daughter!I once did on macaroni and it was okayLater I learnt this was called a 'struggle meal'
>>101997065>I would disown you if you where my daughter!huh?
>>101997140's what I said boy, so don't you go and have any of those 'modern' ideas you hear about on the news and in 'collage', and go change your gender at the DMV or smthing, bc you'll be disowned! out! on your butt!
Can someone give me a QRD on running XL models on A1111? Is it just a matter of downloading and using an XL safetensor w/ A1111? Or do I need to be changing settings? Just tried running an XL base model, and the results look super crispy. I've only been running SD 1.4/1.5 models until today.
>>101997193I'm on comfy but it should work out of the box, load the checkpoint and adjust the clip skip to -2, I believe, or maybe that is only with ponygood luck, post picshave a picture of my pussy, it's really hairy
>>101997175>>101997175>>101997175
I burnt my elf yesterdayShould've turned her on the other side quicker
>>101997367>Should've turned her on the other side quickerthis is good advice, always remember to turn your elf waifus over in the sun to get a nice even tan
omg it migu>>101997367She gon be a Dark Elf soon
>>101997412so that's why their always so angry!they're sunburnt
>>101997323Thanks. Yep, it does work ootb, mostly. Problem was RTFM; model maker recommended particular settings for sampler, CFG and sampling steps. They were set to high, and so image looked fried.
I'm helping
Who's that Pokemon
working on more efficient batch support for florence_toolcuda usage with current batching is up and down because it runs the processor on each batchthis version uses a pytorch dataloader, afaik the limiting factor in this test is windows environment because it can only use 1 dataloader worker, should be higher on linuxanother possibility is because it's running batch_decode then post_process_generation after each batch, i'll do something with that if i can't get sustained 90-100% on linux