Previous /sdg/ thread : >>101138197Actual gen edition (No schizo posting allowed)>SD3 info & downloadhttps://rentry.org/sdg-link#sd3https://education.civitai.com/quickstart-guide-to-stable-diffusion-3https://aitracker.art/viewtopic.php?t=57>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>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
Gens only
Cursed thread of eternal lolcows
>>101145663Comfy thread for gens
>debo trying to stealth post again
>>101145622Can SD3 be saved or all of you too angry by the license and lack of futa? Because from what I've used of it, it does a lot more than SDXL does, already. The Chinese models have a long way to go.
>>101145690Post gen
>>101145698I've heard that it's much harder to finetune than SDXL was, but I'm no expert
>>101145634>>101145673 These are really great.
fill this one first >>101145156>>101145156>>101145156
baking
>>101145724Ty anon>depicting a podcastInteresting prompt btw, nice result
fill the first thread instead of making new ones you mongrels >>101145733>>101145733>>101145733
https://www.reddit.com/r/StableDiffusion/comments/1do5gvz/the_open_model_initiative_invoke_comfy_org/>Together with the community, the Open Model Initiative aims to bring together developers, researchers, and organizations to collaborate on advancing open and permissively licensed AI model technologies.>The following organizations serve as the initial members:>Invoke, a Generative AI platform for Professional Studios>ComfyOrg, the team building ComfyUI>Civitai, the Generative AI hub for creators>LAION, one of the largest open source data networks for model training>To get started, we will focus on several key activities: >•Establishing a governance framework and working groups to coordinate collaborative community development.>•Facilitating a survey to document feedback on what the open-source community wants to see in future model research and training>•Creating shared standards to improve future model interoperability and compatible metadata practices so that open-source tools are more compatible across the ecosystem>•Supporting model development that meets the following criteria: >True open source: Permissively licensed using an approved Open Source Initiative license, and developed with open and transparent principles>Capable: A competitive model built to provide the creative flexibility and extensibility needed by creatives>Ethical: Addressing major, substantiated complaints about unconsented references to artists and other individuals in the base model while recognizing training activities as fair use.>""Ethical""Oh... oh no...
>>101145733Nahh that's literally just one poster having a meltdown for 45 minutes, it's not an /sdg/ thread
>>101145733Hell no, that one has gone to shit already
>>101145763>Thread has existed for 46 minutes>55 repliesReally makes you think. All it takes is one /ldg/ poster replying on cooldown to completely fuck up a thread.
>singular schizo anon theory
>>101145761Same cancerous buzzwords but multinational entities such as Google/MS/etc continue scott-free.Tale as old as time.
>mfw Resource news 06/24/2024>"style transfer precise" Added to ComfyUI IPAdapter plus https://github.com/cubiq/ComfyUI_IPAdapter_plus#important-updates>Stable Cascade was originally MIT licensed? https://github.com/Stability-AI/StableCascade/commit/209a52600f35dfe2a205daef54c0ff4068e86bc7>Geometry-Aware Large Reconstruction Model for Efficient and High-Quality 3D Generation https://github.com/alibaba-yuanjing-aigclab/GeoLRM>VideoGenHub: One-stop library to standardize inference and evaluation of Conditional Video Models https://github.com/d8ahazard/VideoGenHub>AI’s Most Ambitious Music Generators Accused of ‘Massive’ Infringement In New Lawsuit https://www.rollingstone.com/music/music-news/record-labels-sue-music-generators-suno-and-udio-1235042056/>ECLIPSE: Expunging Clean-label Indiscriminate Poisons via Sparse Diffusion Purification https://github.com/CGCL-codes/ECLIPSE>Contextual Interaction via Primitive-based Adversarial Training for Compositional Zero-shot Learning https://github.com/lisuyi/PBadv_czsl>MOFA-Video: Controllable Image Animation via Generative Motion Field Adaptions in Frozen Image-to-Video Diffusion Model https://github.com/MyNiuuu/MOFA-Video>SD.Next release 2024-06-23 https://github.com/vladmandic/automatic/blob/master/CHANGELOG.md#update-for-2024-06-23>ComfyUI Unique3D https://github.com/jtydhr88/ComfyUI-Unique3D>AutoStudio: Crafting Consistent Subjects in Multi-turn Interactive Image Generation https://github.com/donahowe/AutoStudio 06/22/2024>ComfyUI DiffSynth wrapper nodes https://github.com/kijai/ComfyUI-DiffSynthWrapper>Update on SD3 on Civitai https://civitai.com/articles/5840>ExVideo: Post-tuning technique aimed at enhancing the capability of video generation models https://huggingface.co/ECNU-CILab/ExVideo-SVD-128f-v1>ComfyUI-JNodes: Variety of python and web UX improvements https://github.com/JaredTherriault/ComfyUI-JNodes
>>101145761fuck yeah burn SAI burn
>>101145761>retardedtrannyorgit's doomed
>>101145761have some tiny hope with civitai being there but yeah i bet this is gonna be same shit different asshole
>>101138235Oh I just saw this, this was my prompt for that gen:Animal, masterpiece, epic, cinematic, flowers, epic, 16k, anime, (fantasy), soft lightingNothing special, feel free to post the model though that'd be cool
should i join the discord? seems like the place every normal anon went to?
>mfw Research news 06/24/2024>Image Conductor: Precision Control for Interactive Video Synthesis https://liyaowei-stu.github.io/project/ImageConductor/>HLQ: Fast and Efficient Backpropagation via Hadamard Low-rank Quantization https://arxiv.org/abs/2406.15102>A3D: Does Diffusion Dream about 3D Alignment? https://arxiv.org/abs/2406.15020>Disability Representations: Finding Biases in Automatic Image Generation https://arxiv.org/abs/2406.14993>VividDreamer: Towards High-Fidelity and Efficient Text-to-3D Generation https://narcissusex.github.io/VividDreamer/>Multimodal Task Vectors Enable Many-Shot Multimodal In-Context Learninghttps://arxiv.org/abs/2406.15334>LLM2FEA: Discover Novel Designs with Generative Evolutionary Multitaskinghttps://arxiv.org/abs/2406.14917>Demonstrating the Efficacy of Kolmogorov-Arnold Networks in Vision Taskshttps://arxiv.org/abs/2406.14916>LatentExplainer: Explaining Latent Representations in Deep Generative Models with Multi-modal Foundation Modelshttps://arxiv.org/abs/2406.14862>Six-CD: Benchmarking Concept Removals for Benign Text-to-image Diffusion Modelshttps://arxiv.org/abs/2406.14855>Is A Picture Worth A Thousand Words? Delving Into Spatial Reasoning for Vision Language Models https://arxiv.org/abs/2406.14852>Latent diffusion models for parameterization and data assimilation of facies-based geomodelshttps://arxiv.org/abs/2406.14815>Evaluating Numerical Reasoning in Text-to-Image Modelshttps://arxiv.org/abs/2406.14774>Holistic Evaluation for Interleaved Text-and-Image Generationhttps://arxiv.org/abs/2406.14643>Stylebreeder: Exploring and Democratizing Artistic Styles through Text-to-Image Models https://stylebreeder.github.io/>DragPoser: Motion Reconstruction from Variable Sparse Tracking Signals via Latent Space Optimizationhttps://arxiv.org/abs/2406.14567>This Looks Better than That: Better Interpretable Models with ProtoPNeXthttps://arxiv.org/abs/2406.14675
>>101145902ran is stinking his diaper right now
>>101145920i dont even know who ran isis the discord just shit flinging and drama spilling over here? genuinely asking
>>101145761Wow! I'm actually super excited by any new thing that Comfy is involved in. He gets a lot of hate here, but ComfyUI had won over most of its detractors and I imagine he's not exactly keen on cucking the model in the name of ambiguous "eThIcS". This news brings me hope, /sdg/
>>101145935such is the nature of attention-starved troons
>>101145935no, just catjak embarrassing himself attacking ani all the time. honestly think he's the schizo
>>101145965>trannoid schizobabble
>>101145959>but ComfyUI had won over most of its detractorsliterally nobody uses it
>>101145935they are power tripping and want to assign "ranks" to each other based on subjective feelingsthey even force you to upload one of your gens with metadata so they can impersonate you after that to start shitthere is also doxxing, one of the mods said a couple of threads ago they have enough material do doxx "3-4 anons, but why would we"do not join it
>>101145959You could always write your own set of Python scripts. It's not that hard to implement a basic workflow.
... new thread?
>>101145761>LAION, one of the largest open source data networks for model trainingDOALAION are just as responsible for SAI's bullshit
>>101146006Old one is 55 posts from 2 IPs and one single gen. Let it die quickly.
>>101145902I feel like I joined a sdg discord the other day but it's not in my list, did I get banned without saying anything?
>>101146022are you staff? the IP counter was removed months ago
>>101146022>2 IPsah back at making shit up "anon"
>>101146025>seems like the place every normal anon went todo you retard trannoids not realize what the word anon means?
also SAI provides LAION's compute lol
>>101145622test
>>101146039>>101146034It's just speculation but yeah, that thread is fucked, let it die
>trani exposed yet again>suddenly there's a push for a new thread for no reasonwhy does this always happen?
>>101146079Check out this cool gen I made
>>101145761This is lead by the invoke people and there's a good chance it doesn't go beyond the logo stage but a true open model would be nice.
>>101146104thanks thread schizo
>>101146104>he's now embedding CP in genscome on now
>>101146119Are you currently able to arrange any funding for compute?
>>101146160This is you
>>101146161I don't trust invoke to coordinate a good model, this is just for PR I think
>Ani is crying
>>101146181I half agree, but half of me thinks that if they can get enough funding and focus on the training it can't be too fucked. Certainly better than nothing while we wait to see if PixArt will deliver
>>101146119>joins a team and immediately shit talks his peers
>>101146220it truly boggles the mind
>>101146203why?
>>101146119Agree, invoke have been all talk before, they already got funding and have nothing to show for it
>>101146119
>>101146274>have problem>post in thread>dev fixes itThose were the days man, fuck
>>101146161There's not even a model arch that's decided on yet.>>101146220Being honest is shittalking? I don't want to hype this up until there's something concrete.
>>101146293nah having garbage like trani around is better
>>101145846>>101145918thanks for getting some news up. sorry I missed the timing this morning>>101146203are you really going to put in another 16 hour shift of this shit? >>101146018did laion ever republish its dataset? last I knew they took it down but never heard of it ever going back up
>>101146327Still less hours than you
>>101146327>are you really going to put in another 16 hour shift of this shit?we learned from the best
>reply to me with your gen and i will only respond if its one ive seen before so i know who it is im talking to >on an anonymous chinese shill website even
>i wasnt even here guise whats habbening
>>101146302Is it true that the model might be trained only on synthetic images?
Here's some good news on the open model, one of the first people the invoke people invited was the pony model creator so things might end up being the complete opposite of safe.>>101146408Hopefully not. I think training on synthetic data is dumb.
>>101145156>>101145156Stop flooding
>>101146426>https://www.reddit.com/r/StableDiffusion/comments/1do5gvz/the_open_model_initiative_invoke_comfy_org/this should be posted to comfy.org
>>101146457This is you
>>101146426when you hear the shit astrabitch says you will change your mind
>>101146483What do you mean?
>>101146327idk. i think 5b has been dropped for datacomphttps://github.com/mlfoundations/datacomphttps://www.datacomp.ai/
why the thread splitting shit again? why dont you fill up the other one first?
>>101146520the mysterious "baker anon" (not debo) decided for us because he thinks its his thread
>>101146520they like to imagine they hold some kind of power
>>101146541this is why the pastebin is needed
>he's spamming again
schizofest time?
>>101146620(t)ran is shitting his diaper hard keep going
>>101146645at least he's not a pedo
>>101146605cool
>>101146659>faps to toddler scat>not pedoit's always the ones that deny it the hardest
>>101146679post a proof and i will hate ran too, promise
>>101146665Ty anon
>>101146694yup is holding out on us but it's real. trust me
>>101146713so no proofthen its only trani confirmed who deserves all the hate, gotchu
>>101146733There's a ship in this one
>>101146733>ran "anon" post detected
>>101146713>>101146746pathetic
I don't know if anyone remembers me but I'm the anon who was posting about his bolognese lunch yesterday. Quick update: it was really good, it's always better the day after I find. My secret ingredient is a tiny bit of Worcestershire sauce
>>101146746might sound strange to you but a lot of people hate pedos and no im not ran
>>101146761Based
>>101146761Bolognesechad strikes again
>>101146772I hate ran and ani because of that
>>101146843Any tips on getting a good zelda gen?
>>101146713mental illness
>>101146895not really. XL seems to know her well enough. Just make sure to have princess zelda and ass in your prompt and link in your negatives
>>101146913>link in the negativesgenius
>debo and trani trying to shit on ran againits not working
>>101146959v cool. Is there a percentage of mobius in there?
so we are stuck herei wonder what SAI thinks about one of their employees acting like shown here in public, mhhhh
>>101146973Yes, probably even a (moebius)
>>101146959>>101146973obvious samefag is obvious
>>101146981meds
>>101146994doh>>101147000kys yourself, schizo
mhhhh lets find out then i guess
>>101147000This is you
>>101147020based schizoanon
Good thread of goodness!
Morning anons
>>101147254be careful or trani grooms you soon
>>101145761>unconsented references to artistsI'd like this to be addressed ASAP. SD3M is impossible to get any type of style by reference to artists, neither living nor dead, and for this reason it is much worse as a model. My complaint about SD3M goes beyond not being able to copy an art style, it is very hard to "move" the style in any desired direction.By how this principle is written I don't have high hopes about this project: obviously dead artists cannot consent, and living ones I don't think that they are going to be in a rush to give consent.The truth is that any artist can copy an art style from another artist, but it seems like AI is not allowed to do so.Dead artists cannot consent,
so i got auto1111 with lora working i think, but it's not using my vram. any way to force it to use my vram. In auto1111 it's setup to use vram but it's not?
>>101145622When are they going to update nvidia drivers for linux? It cant even do SDXL where it locks up my computer and I have to do a hard reset.
>>101147262Morning
smashed out the rest of unetsa lot of autoencoders, transformers and unets are made up from base modules and most of the base modules are covered by unittests so hopefully everything should be goodmissing ops will be easy enough desu
>>101147268way ahead of you, anon!
>>101147277That's a better Lara Croft than what the DEI people cooked up for her anniversary gamethat one looks horrible
>>101146772>ran detected
>>101147459>trani hunting new victims
So 1.5 + NAI + lora is the best model?Because it seems is easier to control than XL.
motherfuckin debo with duplicate op
more clones than i expected desu
>>101147601SDXL is pretty good, but 1.5 is still king in various aspects.SD3 fucked itself senseless so outlook is incredibly grim
going back to old ass schizoprompts lol
>>101147838Oh my
>>101147539Super cool btw
>>101147838I remember those, it was fun but definitely harder to remember.
>Ani is afraid>Ani is leaking shit from his torn asshole
>>101147947>Ani is afraidrightfully but its only going to get worse from now on :)
Do midjourney and shit still inject "black man" into all its prompts or are we past that phase?
>>101148018I think it's time for all the truth tellers to unite :)
>Julien is crying
>>101148187nice
hey, do you want to see my waifu?
Julien
>>101148613Reclining on a Chaise-longue
imagine being a lolcowon 4chan
>Pajeet new CEOWhy?
>a lolcow
>>101148862They run the world, brother
>>101148869you mean they ruin the tech world, brother
my A1111 with direct-ml just fuckin crashes and exits every time i try to generate something, it just hangs
>>101148869Sar I am microsoft, you're computah has teh birus, I need yuu to go to teh Taaget and get teh $500 gift kaaad maam
>>101148975lol good luck with no more competent devs
>>101149095what happened?
>>101148975Yeah I had an old install from a year or so ago, tried to update then reinstall and it just shits the bed immediately. ComfyUI gave me cancer, seems like invoke works though.
guys guysim literally shittingright now
so sleepy. maybe i can make a start on the compilation side of things tonight though
>he's still crying
>>101149120Julien had that problem since he was 14
>>101149120>>101149186good news, this time it went well, the fabled ghost wiperi still used tp but it remained clean
Everything you guys share is mediocre shit. Stop making these threads. Nobody cares. Go outside and lay in real some grass for once in your lives.
How does lora merging work? Same way as models?
>>101149335Cry scrub
>j
so how did the pony xl guys make a model that eclipses sd3 for anatomy?OH RIGHTthey trained it on literally everything with no censorship. (this is autismmix, a checkpoint based on it)
>starts posting here again
>>101149421and how do you get a good model? good training data.if sd3 is censored, it's trained to inevitably fail at basics like figure drawing. meanwhile pdxl is trained on everything, so it doesn't.
>>101149421>>101149459who's this weirdo even talking to. You can't sit with us, anon.
>>101149519im just a random anon posting gens. I dont know who "us" is.
>>101149536>posting gens.gross, this is a schizo nogen thread
>>101149536can you just stop?
>>101149421No censorship, until the very end where the brony author went like "oh yeah, let's remove all these tags because maybe artist will like us more then"
>>101149536Ignore the nogens. You can post what you want.
>>101149628They got to him
>>101149536I heard there's a discord with open invites that would love to have you, just need to post an image of your government issued I.D. to be verified
is jules still buttmad? c'mon, i (schizo anon) was just joking!
>>101149704Ani is still crying
/shitty drama general/
>>101149741Alternatively: /Shizo Delusions General/
>>101149644cute alpaca, model?
>>101149851NovelAI. It's the best for characters currently, which is unfortunate.
>>101149868oh true this is not the local image gen thread, i forgot, my bad, bye
We need some whimsy to lighten the mood around here
>>101149965Very creative
Is Segment Anything still the meta for automatic detection and masking?
>>101149965>>101149879These looks like the things 50+ yo women like.
I can't tell if this is SDXL not understanding how pillows work or SDXL understanding exactly how people use pillows sometimes.
Who do we hold accountable for killing this general?
>>101150477The nogens thirsty for drama and obsessed with namefags.
>>101150499t. namefag angry that anon wised up to his games
mm definitely shoulda been more careful with that mug
>>101150477time and fate
its pw's birtday today and they're not even hanging out with us
>>101146160How? I thought metadata and stuff like that was stripped from image uploads
>>101151040hide a gem
>>101150951>>101150981Cool it with the interracial stuff
>>101151148These are great.
>
>>
can get some good results in pdxl without a lora even (this is the nai3 model, autismmix also works well for anime)just use basic booru tags, this works amazing for it (any model, good for tags just type a word or show and you'll get ideas)https://github.com/DominikDoom/a1111-sd-webui-tagcomplete
>>101145892>have some tiny hope with civitai being therecivitai owners don't like porn, just so you know
I was searching google to try to find info about SDXL full fine tuning learning rates, and lo and behold, the #2 result is some fraud's AI generated blogspam article. It's fucking over. The internet is being overrun by AI slop. There still is good information out there, but it is becoming increasingly difficult to find. I know people have being complaining about this for a while, but this was really the first time I genuinely needed to find some niche piece of information and the AI generated spam shit makes it near impossible.I did eventually find a good source, btw, thanks to a fleeting memory I had of seeing it months ago. The Animagine XL 3.0 model card on huggingface says what all the hyperparameters used were, and looks like a good starting point for anyone doing a full fine tune of their own. And of course google will never find it when searching for SDXL hyperparameter best practices.
>>101151541>LAION already stepped out>Astratroon was the first finetunes they approached>Emad wants inDOA
>>101151574theres also this fine tune, it has the hyperparamshttps://civitai.com/models/351170/animeboysxl-v3
Good afternoon, anons! I hope everyone is doing well :]
>Avatar or signature use
>but its not an avatar i swear
>>101151638>>101151650Get a grip. No one cares.
@101151663>nig*bo
>>101151541Really? well they picked a bad business then judging by any casual perusal of their hosted files
>>101151627Thank you, not bad I guess! Drama filled threads though... Or the usual.
>>101151623Thank you, gay king. Seems like everyone tunes the text encoders also, even though kohya documentation recommends against it. I'm currently doing a unet only tune, but from past experiences with loras, unet only struggled learn concept-style data. More things to experiment with I guess.
>>101151729Glad to hear that, anon! :]LOL yeah that's unfortunate hahaHopefully not too much af that goes on today
>which persona are you going to get today
imagine not replying to her with an image. how will she know whos talking to her on an anonymous basket weaving forum?
>>101151895psycho fuck
@101151905nig*bo rattled
>hm?
Modelbashing Pony and 1.5 is fun.
>>101151627hi pw. happy birthdaystay out of trouble! or get into trouble... whichever sounds more fun
>sd3 cant do the girl on grass prompt>pdxl can do it just fine with many characters even without a lorastability ai, your excuse?
>>101151870Good afternoon, Fran! Long time no see! :]It's so great to see you again!>>101151937Heya Debo!! :DThank you so much!! Hahaha it's still early so we'll see! I hope you're having an amazing day so far! Cute PW gen!!
>>101152066Happy Bday
>>101147308>The truth is that any artist can copy an art style from another artist, but it seems like AI is not allowed to do so.this
>>101152066Sup, nice blond chick. Have you though of genning any other types of witches? I think a red fire witch would be pretty cool.
>>101152094Thank you so much, anon! :DI hope you have a great day! Really cool gen! I like your alien haha
>>101152128Thanks PW, messing with Pixart-sig and upscaling in a SUPIR workflow. Have a good evening!
>>101152122Sure hahaha here ya go!>>101152161Oh cool!! I haven't tried that yet hahaI'm playing around with SD3 right now :]Thank you so much! I definitely will!
>>101152199Nice! Also happy birthday
hnnng
>>101152227I'm glad you like it! :DYou too! I like your witch as well :]Thanks Fran!!
>>101152199I like SD3, it needs love. So much hate directed at the model, there's a fine line and the passed it making the model sketchy. Does the best eyes/lips hair! etc
>avatarfagging intensifiesI'm out
>>101151541>don't like pornThey found a different hosting site though that allowed their uses to gen porn when the original one they used disallowed it thoug
>>101152382I agree! It's great at what it does, especially recognizing your proompts! The only real issues i've noticed is that hands come out funny, and poses aren't as dynamic, but everything else looks great!
>>101151541good, fuck coomers
>>101151595LAION is already out? Lol
Now that SAI is officially dead, what is next for Stable Diffusion?
howdylooking at old prompt files today
>>101152487happy birthday pw! <3 I should make something for the occasion!
>>101152646
>>101152710
the good ol' days>>101152733
>>101152661Animanon!!! :D Hello!! <3Thank you so much!! I'm so happy to see you :]I hope you've been wellGreat animation!!
>>101152758
>>101152764happy bday PW
>>101152767
requesting a bake with a pw captcha has pw wtf?PWDYD
>>101152764>Thank you so much!! I'm so happy to see you :]same! you always brighten my day! I think this one will end up really cute! you should join the discord server too!
>>101150941>>101151627happy b-day, have a witch >>101152808
>hand becomes fabric>fabric becomes handyep its sd3 alright
>>101152931
>>101152962That doesn't look very good. Why post it?
>>101152986I'd be too embarrassed to post without a gen. imagine how cringe that would be
>>101152964
bake?
>>101153018it's ded, jim
bakers?
B A K E RA K E R BK E R B AE R B A KR B A K E
>>101153168i could but it's going to tell me to wait 5minutesif no one has by then i'll click 'submit'
>>101153189>>101153189>>101153189
>>101147824>SD3 fucked itself senselessWas it ethics?
>>101151760>even though kohya documentation recommends against itthey just say the results will be unpredictable because they didn't test it, but there's no problem in training them when fine-tuning, from what I recall if you are training a new word (e.g. a character) training the text encoder is almost mandatory, whereas with stuff like styles you can do with unet only, though it will be a lot slower with unet only training
>>101145761ensure to overwrite all tokens,commonly used words and slang, that describe minors - overwrite it with hansen tv guythis ensures anyone who is a dumcunt will have to explicitly and deliberately train model to produce such imagesif they do so, they might get proper help then