Goo' editionPrevious /sdg/ thread : >>92183526>Local installNvidia GPU: https://rentry.org/voldyAMD GPU: https://rentry.org/sdg-linkCPU: http://rentry.org/cputardAlternative UI (Node-based): https://rentry.org/comfyui>Run cloud hosted instancehttps://rentry.org/sdg-link>Try online without registrationtxt2img: https://huggingface.co/spaces/stabilityai/stable-diffusionimg2img: https://huggingface.co/spaces/huggingface/diffuse-the-restInpainting: https://huggingface.co/spaces/fffiloni/stable-diffusion-inpainting>Newshttps://rentry.org/niakonichan>Models, textual inversion & embeddingshttps://civitai.comhttps://huggingface.cohttps://rentry.org/hdgrecipeshttps://rentry.org/embeddings>Wikihttps://www.sdcompendium.com>Index of guides and other toolsWant to know how to use SD for animation, upscaling, train your own model or merging model checkpoints?Or need guidance on prompt ideas and when to use which artist then follow these links...https://rentry.org/sdg-linkhttps://rentry.org/rentrysd>View and submit GPU performance datahttps://docs.getgrist.com/3mjouqRSdkBY/sdperformance>Share image prompt infoIf you would like to easily share your image generation settings (prompt, model, etc), since 4Chan removes that information from the EXIF, then follow this guide https://rentry.org/hdgcb>Related boards>>>/h/hdg>>>/e/sdg>>>/d/nai>>>/b/degen>>>/vt/vtai>>>/aco/asdg>>>/trash/sdgOfficial: discord.gg/stablediffusion
1girl
can you get banned for a sfw image where the thumbnail looks nsfw
>>92185464Bump
>>92185481oh manif only earth was this awesome
>>92185464Which model did you use?My realistic models can only make asians
>>92185490what is the purpose of being such a shitty fucking poster
>>92185502What do you mean?
emma
>>92185495you never been at the grand canyon, right?
>>92185513well fish tanks nobody gave cellphones a chance all the times things going yesterday?
>>92185496Just don't use lofi or chillout, those are heavily asian. But even on those you can just negative prompt (asian:1.2), (japanese), (korean) and you'll get far less.
>>921854643girls
>>92185518thats not emma.. this is emma
>>92185522I can mention other places that are cool toobut they fall short of the earth that I want
>>92185526Yes
>>92185473>>92185451overall I find structural and chromatic coherence, detail and clarity on average to be better at higher samples in euler a.
>>92185542looks more like 3 kot
>>92185563Well then go for it bro, literally the entire community agrees on this subject. But go for it.
>>92185558flabbergasted crumpets gallivant across the perpendicular rainbow of inverse flamingos, while the xylophone whispers to the wind
>>92185565yes
>>92185581most of the community regurgitates eachothers prompts with little to no original input using eachothers loras and chekpoints to avoid doing any work at allwhy would you assume they would behave any differently to observations about outputs?
>>92185583yes
>>92185452I posted this earlier and some anon was trying to help me but my internet got disconected:I can upscale with img2img just fine, but txt2img I just can't with anythingIt's weird because I have 11gb vram, I should be able to do some thingsbut every time I try I get the same results:it goes up to 50% and then it stops and it outputs a blurry unfinished upscaleNow this anon said to use 1.5x at 0.3 denoising.And it worked for some reason.So I said to myself well that's nice, it was the denoising. It might take a lot of vram.So I put it at 0 and same error.For some reason I can only upscale in txt2img at 0.3 denoising, not more, not lessAnyone know why this happens? I'm more curious about finding a reason than a solution
>>92185602>most of the communitycommunity? what is this "community" you are talking about?
Raita becomes infinitely more tolerable if you mix his style with something that doesn't have torpedo tits
>>92185534Which one do you use? I don't know any others than lofi, chillout and uber realistic porn merge
>>92185624my brainstopswhenI haveto readthis
>>92185624based on the pic, maybe your problem is faggotry, stop being such a faggot and it will work
>>92185627i responded to a post mentioning the community
>>92185602I'm not talking about fucking randos on 4chan like you. This has been tested since day one with ancestral samplers in SD and even if we hadn't tested it, a basic fact about how an ancestral sampler works.Please fucking google. Don't ask questions if you don't want to listen to the (correct) answers because you don't want to hear that answer.
>>92185640Okay from now on I'll stop. I just thought it was an easy way to digest whatever I say
>>92185624you have next to no available memoryhow many crutches are you leaning on to get the results you want?
>>92185657>reee I am better than you 4channers even though I am right here with you>reee WE tested it and know better than YOUgrow up
>>92185663Crutches? Exactly what do you mean?This happens with any prompt
>>92185681post the settings of an example prompt.. hard to analyze anything of your problem with no data
>>92185681your conda is fucked, reinstall all of that shit
>>92185681Your setup is claiming way too much memory by default.
>>92185638Deliberate, level4, or a custom merge of delib+lofi+level4.
Can you refine/train an existing LoRA to further improve it or do you have to start fresh each time?
>>92185689Here you go in pic related. The image resulting stopped at 50% again with 0 denoising. So it didn't upscale correctly. I just tried again at 0.3 denoising and it worked correctly. Which is what has gotten me confused.>>92185691Alright no problem, but first could you explain what conda is?>>92185703Why could that be?
https://rentry.org/ezloraneeds work but the bare basics are down
>>92185678That's what you took from that?I think you'd see that pretty much everyone in these threads would agree about what I said about ancestral samplers. Just move along and keep chugging out your 50+ step trash, wasting time and resources for your placebo improvement.
>>92185735there's definitely something fucked with your installI can make shit in much higher resolutions with 6GB VRAM
>>92185735>high res fix: on>denoising strength: 0 (zero)>euler aI can see a few problems, the biggest being you didn't read the manual. the second being you are not using new, faster, samlpers.
>>92185735first of all, denoising strength zero is totally pointless, without any denoising hires fix doesnt work.. and a setting of 0.6 is about standard for most applications (unless you doing img2img) .. second of all your setting seem fine other than denoising, id do a fresh install, just make a new folder and do clean slate.. something in the install process might be borked
>>92185778>high res fix: onyeah that's how you upscale genius
anything I've missed out on this month
>R_U OAO released, buy my new album
>>92185735just reinstall the whole thing
>>92185789>curiosity killed the catnot really.. welcome back brazilo gato
>>92185788Stop talking idiot. High res fix on with denoising set to 0 doesn't do shit.Who the fuck are all these newfags in these threads? First that 50steps ancestral idiot now you.
>>92185805stfu, do ya wanna fight wit me huh? i see you in the ring, boya, a box match right know, i will knock your ass down mf, come one man
>>92185534>>92185638I use lofi, just use (russian face) or (american face) that usually prevents asiansmaybe (award winning) does sth as well
>>92185805That's not me butStop being immature. Insulting others over your own ignorance won't make you happy. I'm just trying to learn here. You, I don't know what you are trying to do but you are doing it wrong.>>92185779>>92185795Will do. My point was that the upscaling worked only at 0.3It didn't work at higher or lower. 0 might have been a poor choice and the conversation was derailed.
>>92185838blonde hair is also much more likely to give a white looking face, unsurprisingly
>>92185624Hi anon, I was the person earlier who suggested that. Are you using a laptop with integrated graphics? If so, I have bad news...If not, which GPU do you have?
mood
>>92185868not that anon, but 11GB is probably a desktop 1080 Ti
>>92185838Are you saying asians never win awards?
>>92185868Hi, thanks for the hand earlier.Luckily it's a desktop, it's a 2080ti
>>92185880>10802080
>>92185882not for beauty noI think most miss universes and shit are some South American mulattos no?
>>92185848Are you serious right now?Did you actually read the conversation?The dude smugly corrected someone by calling him a genius while also being completely wrong.This is not reddit faggot, shut the fuck up.
>>92185910>The dude smugly corrected someone by calling him a genius while also being completely wrong.where did this happen?
>>92185900Oh, yeah that sounds plausible actually
>>92185910Chill dude, literally nothing for you to earn for being overly aggressive besides coronary disease.How was he wrong? I thought you needed to use highresfix to upscale
>>92185885Questions:1. What resolution are you starting with?2. Which upscaler are you using?3. What multiplier (x2, or dimensions) are you trying to upscale to?4. Do you have xformers on?5. What "batch size/batch count" are you using?
>>92185788its not my job to read the manual to you fren.zero denoise doesn't do ANYTHING. thats your one free tip. now go read the manual.
>>92185935Holy shit is this literally your first time on 4chan? You see those little numbers you use to quote other posts? CLICK THEM AND FOLLOW THE CHAIN.
>>92185965this guy wildin'
>>92185961I don't think that guy was trying to do anything other than make you upset leave it.Do you mean it doesn't even upscale? I mean I did read that at 0 nothing would change. But I thought that referred to the image composition.
>>92185942I'm not even going to answer that as it has been answered 4 times already why the COMBINATION of settings is wrong.You seem to not understand how basic 4chan post/comment chains work.
are you all just going to keep arguing instead of giving me (You)s?
>>92186023i don't even know who (YOU) are anon.
>>92185998stable diffusion works by making a noise pattern and for lack of better human terms bluring it together into "what you want"when you upscale it has to noise the image you already made and denoise it into the upscaleif you set it to zero it won't do ANYTHING.if you set it low in img2img (0.03) you can make images look almost exactly the same as the original while upscaled, but if you do it wrong there will be artefactsI hope you learned something today, I had this problem once. please be patient, its not easy to pick up all the tricks in one day.
a thread full of trolls, newfags, and basically a tech support general for one guy who still refuses to read the manualgood thread yall
>he's still seething
>>92185989that would be japanese moot.
>>92186049>no picture
>>92186049just another day on 4chan, boss
>>921859501: any resolution has the same results, but let's say 512x5122: any upscaler has the same results too, but lets say latent for the sake of practicality 3: usually x24: no5: I don't touch that, I am using just 1 at a time>>92186009Damn dude, having this kind of mindset can't be healthy. I mean if you didn't want to answer it you wouldn't have answered lmao. Not trying to fight here, this conversation seems to make you angry and I know that is unpleasant.>>92186047Thanks man, I appreciate it.
>>92186061you will never amount to anythingyou have no social lifeyou have no girlfriendyou have no job prospectsyour only enjoyment in life comes from trying to make other people miserable, because you are miserable
>>92186049If my problem was covered by the manual somebody would have already pointed to a solution, though.It isn't I checked.
>inpainting faces into 6k imagesthis is the futuremy webui almost crashes when I do it but whatever
>>92186068I've probably posted 25% of the pictures in this thread, I don't want to bore people by posting the same type of content from 2 of my batches over and over
>>92186089just segment the image first, smartypants
>>92186103hmm? you mean like manually split it and then recombine? seems like too much works unless there's another fancy pancy extension for that
>>92186049all me!
what's a lycoris
>>92186098>I've probably posted 25% of the pictures in this thread,delusions, take your meds
>>92185723I would still like an answer to this. I've looed through the guides I could find in the OP and it seems like they are all based around creating a fresh LoRA from a dataset + tags but I feel like it should be possible to just continue training an already created LoRA in the same way that you already do with epochs. Under the list of arguments when training I see "lora model for resume" with the following description:>Loads a "Hypernetwork" into the training model, also works for LoRA, which is why I have it here, this is not the intended way to continue training thoughIf that is not the intended way to continue training it suggests that there IS an intended way to continue training but I don't think anywhere else on the page indicates what that is.
aaaaaaaaaaaaaaaaaaaaaaaaaaaah!
>>92186123>seems like too much work>to cut an imagengmi
>>92186132neat
>>92186154it works fine without doing it just a bit sluggish. If I start generating 8k monsters for some reason I might consider it
>>92186131it's a combination of locon and hadamard product adaptation.locon is a generalization of the 2D low rank adaptation of loras by also applying it to the 4D tensors that handle convolutions in the denoiser.hadamard product is intended to reduce the rank needed for loras. A 12 loha can theoretically capture the same information as a rank 128 lora
>>92186071Which part does it fail on? Is it in the initial creation of the "before hi-res fix" image, or after it's halfway done and creating the hi-res image?Also, assuming you're using Automatic1111 webui, open your webui-user.bat file in Notepad, and on the line that says:set COMMANDLINE_ARGS=change it to:set COMMANDLINE_ARGS=--xformersthen restart it completely. Your images will generate much faster, at the cost of "inconsistent" images (ie. you won't be able to create the exact same image twice, but only with very slight differences, usually).
what happened to comfy? i came back and it doesnt run now :(
>GOO! Love. The Motion Picture
>>92185632I disagree. The torpedo tits are the charm.
>>92186192It fails right at 50% after clicking the generate button when applying highres fix
Is this the right place to ask for help with bullshit? I suddenly got this:>csv.Error: field larger than field limit (131072)Stupid shit, when it worked just fine like an hour ago. How the fuck should I fix it
Windows Defender just decided that the AOM3 .vae is a virus despite me having it sitting on my computer for over a month now.
>>92186190I did not understand a single word of what you just said. Thanks for your time though.
>>92186270could you take a screenshot of your settings on a failure? from the sampling method all the way to the bottom of the hi-res fix settings.
>>92186270This suggests that you can get the original 512x512 just fine but when you are rendering the second upscaled image you are having problems. Have you tried significantly lower base resolutions? For example trying to create a 256x256 and upscale by a factor of 2?Reading your initial problem I would be inclined to say there was just a glitch in your install or something. No logical reason why it would work only on denoising of 0.3 and not higher or lower. For the sake of curiosity, how precise is that 'only 0.3'? If you try, for example, 0.31 or 0.305 will it work?
>>92186281which ui are you using, comfyui or automatic1111
>>92186342automatic1111
Comfy life
>>92186344try a completely empty (no positive nor negative prompts) generation and see if it works
>>92186266
>>92186369The UI doesn't start up at all. The .bat just kills itself.
is midjourney secretly based?
>>92186427>is midjourney secretly racist?yes, and you're probably too
1girl dancing
>>92186440>secretlyDoubt
>>92186308Ah, that just happened to me too
>>92186401I bit the bullet and reinstalled it, now it fucking works.
>>92186445based dep
>>92186460If you're in the same boat as the two other Anons maybe Windows decided to delete an important file for your own safety.
>>92186401open your stable-diffusion-webui folder, right-click, select Git Bash Here, and type:git checkout master
>>92186401So, what message appears on the terminal?>>92186308File a bug report on their github. I had to do that last year when it triggered on a well known test framework.
Hello, trying out comfyUI today.Is it possible to share the models folder with a111?
>>92186458>>92186308pickle
>>92186315>>92186334xformers fixed it, I can't upscale certain things like at x2 but I can now upscale some things at leastthank you guys
Glaze BTFO
>>92186507Hey just as a question, I notice that in >>92185735 you have an SD VAE dropdown menu next to your checkpoint dropdown. I don't have that and need to apply VAE from the settings page and I think I can only toggle it from the settings which is obviously less than ideal. How do you setup your dropdown like that?
False positive?
>>92186507xformers is magical, anon.
>>92185961where's this manual?I am new and can't find it in the OP
>>92186551Presumably - Windows Defender seems to have started flagging vaes as Trojans - been hit on both the AOM3 and AnythingV4 one (basically the two I have tried using today).Won't say it is a false positive for certainty, though.
It amazes me how well loras translate the artist style
>>92186551The pickles have sprouted.
upscale: 1.8x 4 batches 0:36upscale: 2.0x 9 batches 1:26
>>92186589https://github.com/AUTOMATIC1111/stable-diffusion-webui-feature-showcase
>>92185161>another model deep floyd is coming out soon tm'member distilled?'member IF?soon:tm: (never)
>>92186490yeah, try this https://github.com/comfyanonymous/ComfyUI/blob/master/extra_model_paths.yaml.example
>>92186550I honestly can't remember. I'm sorry. I used a tutorial pretty sure. It involved making a new VAE folder in the \StableDiffusion\stable-diffusion-webui\models directoryThen after that you had to open the UI, go to the settings tab and do this on pic related.Sorry I can't remember correctly I'll try to find the tutorial
>>92185640Subhuman
The program just trooned out on me!
>>92186604Thank you anon for bringing to my attention the fact that a raita lora exists
What is a good model for a watercolor landscape (or the Disco Elysium LORA)?
Where i can find working torrent with stable diffusion model?
>>92186691Any anime model does watercolor really wellDunno about semirealistic ones. Try dreamlike photoreal
I love maids
>>92186691I would recommend 2.1, but that's a controversial opinion, also for some reason 2.1 doesn't work with ComfyUI very well
>>92186696why would you need to torrent it
>>92186724Wow. Beautiful. Nice.
>>92186654Ah, it was just the Quicksettings. Thank you.
>>92186745Cause stealing it would be illegal.
>>92186744prompt/model? That looks dope.
>>92186680Raita deez nuts
Is there an extension that saves all the latest prompts if I accidentally close the tab or need to reload it?
>>92186764Just go to huggingface.co or civitai.com and download one of the freely available ones.
>>92186722w-where's her hand going??
How do you control the way models are mixed more? Can you, say, mix A (0.5) + B (0.3) + C (0.2)?
>>92186782check the params.txt file in your stable-diffusion-webui folder.
>>92186827you do math.Mix B(0.6) + C(0.4) then mix BC(0.5) with A(0.5)
>>92186772>model?he literally wrote it 2 times
>>92186852Makes sense actually. I just wish there was a more convenient way to do it with 3 models at the same time.
>>92186745Because my downloading speed with direct mehtod is very slow
>>92186882oh i love her delightfully little smug face
>>92186445I had a stoke.
>>92186942I stroked my cawk.
>>92186942>stokeyes, clearly. I'm sorry.
If my dataset has images with odd dimensions will it train properly on them or do they all have to be 512x512 for better results?
When will we get AI smellovision for prompts?
>>92186976You need to crop them using some online image batch trimmer.
>>92186976use a dataloader that supports multi aspect ratio
>>92186910one nostril. i want to poke it,
>>92186976>>92186984It depends on the training script you useMost have 'bucketing' which allows them to train regardless of aspect ratio or resolutionAt the end of the day you could try a Lora with cropping and try one without
What are the prompts to get rid of east asian looking faces? I'm using deliberate v2
>>92186978Quite pungent my dear...
>>92185464please tell me how to generate this
>>92187009just put asian in negatives or type nationality that you want.
>>92187004true, I've tried lora without cropping and it worked to some degree but the results aren't very good, I wonder if it's because it crops them wrong
>>921870451girl, nigger
>>92187045does not work for me
>>92186998holy shit, anon, i didn't even see that lmao. have a soviet hero off-brand Luis Guzman as a reward.
>>92185452I'm reading the manual. How exactly do I do this, with x/y/z plot?Can't figure out how if it is supposed to be like that.
>>92187069That looks like the result of using different weights on the same prompt.
>>92187054use this model instead than i don't know how deliberate works.https://civitai.com/models/17449/level4
So anythingv4 or novelai for style training?
is there RLHF for stable diffusion?
>>92187069>>92187082I figured out all I had to do was look for sampling through the xyz sliders.I'm a dumbass
>>92187106Novelai always
>>92185452Could an anon try and create something replicating the aesthetic of Planescape: Torment?
>>92187036use anime model to generate big tittyuse img2img with lofi at 0.65 denoise or if you want better quality but slower then controlnet + 0.8/0.9 denoise
4girls
>((masterpiece,best quality)), just like some kind of hot monster woman, I dunno, with like fangs or somethingThis shit is easy, bros
>>92187243
>>92187293please tell me what model and stuff you used to make this
I can't stop cooming
>>92187243do you have a NASA computer?
>>92187311anon don't post that, I've bullied my dick enough lately...
>>92187311Based
These threads are always filled with girls, and have no cute boys.
>>92187353kek
>>92187351>>92187353I just can't help myself
random question how do i update the webui i downloaded it a few months ago but i heard they added stuff
https://huggingface.co/spaces/damo-vilab/modelscope-text-to-video-synthesis
Just in case this isn't old news -- LLaMa at homehttps://cocktailpeanut.github.io/dalai/
>>92187310It's a custom model I've been working on, it still needs to cook.
>>92187383Very nice. Model?
>>92187380Re-install or "git pull'. Keep your models and add-on data files backed up so you can copy over the where they are supposed to be if you eff up. You can also keep a copy of the original setup if you really eff up.
>>92187344Haha, no just a 3070Ti
>>92187372rare fellow hair over eyes poster
>>92187372sigh... *unzips*(catbox?)
>>92187406First pass is made with realbiter, second pass through JimEidomode. Still figuring out how Comfy works
>>92187402I heard about this the other day thanks to videostldr PSA: LOCAL text generation is about to get as good as SD.I will be waiting patiently with my 4gb vram card...lol...
>>92187447Oh, so it's comfy. It does seem to produce really crisp and detailed gens from what I've seen of it so far, but I can't really be bothered to switch.
>>92187415hair over eyes is VERY based>>92187425https://files.catbox.moe/6ldkwc.pngA catbox isn't going to help much though. I'm using a lora I trained as an experiment, with a pretty small dataset so it's pretty shit
how do styles work
>>92187481just curious about your prompts mostly, I wanna try doing something like that with the model/embed I usethanks
>>92187243>rock dragons secretly sniffing the butt
>>92187466>tldr PSA: LOCAL text generation is about to get as good as SD.Jesus fuck whatAre we really getting our catgirl wives we can play games with?
>>92187553You can already kind of do that with pygmalion. It's not perfect, but today I got pygm 6b running on a local install with just 8GB VRAM.
>>92187447What denoise amount you use for second pass?
https://civitai.com/models/21510
>>92187604why
hat hat hat hat hat
>>92187604fuck off mongloid
>>92187243what model sir?
>>92187672nice
hey guys, what mask blur level do you use for inpainting? sometimes, the inpaint area looks a little blurry, and I don't know how to fix that..
>>92187391shhhhhh! the queue is long enough as it is!
>>92187687
>>92187686g4ys3xv2.0, do you want a boxcat?
>>92187582I think it was .5
Pixel Art script is kinda kino but also kinda trash. At least it always makes hands look good kek.
>>92185518Emma is made to be the french maid!
>>92187697VERY nice
>started training loras overnight just to keep up with the backlog so it's finally shrinking a bit>update preview pictures>takes like 6 seconds to refresh>what the fuck>check lora folder>95gbjesus christalso on that note, please tell me there's a way to filter networks to only things with previews, this shit is getting ridiculous
>>92187727agreed!
>>92187693it's down
As far as I can tell, two identical mixes made on different machines can never produce identical 1:1 results.Anon here posted a catbox and told me how to make the custom mix. And here are the results.
>>92187735what's the model?
accidentally made a prompt with my inpainting modelwhich style is better?inpainting model is the one on the right
>>92187764anything 4.5
>>92187686https://civitai.com/models/7371/rev-animated
>>92185605make more. I will make chatbots out of this stuff
>>92187749y
how the fuck do I inpaint eyes
>>92187816You inpaint the entire face at full resolution
>>92187816Inpaint the whole face, not just the eye area.
>>92187767That works. You are an artist.
>>92187816Those are perfect, anon
>>92185452
>>92187816lower denoise
>>92187834sugoi
>>92187835I don't get it
>>92185545>I can mention other places that are cool too>but they fall short of the earth that I wantBased. I share the same feeling anon.
>>92187835>Rusty nails scattered on this one road are already working! Carfags will soon fuck up all of their tires and there will be no more cars!Says the horse carriage driver, as he giggles menacingly. Circa 1889.
>tfw leaving your cpu working overnight and finding more than a hundred of incredibly aesthetically pleasing pictures in the morningThat's it. I'm buying an expensive videocard even though I don't play vidya.
>>92187878Why would you generate a hundred identical pictures?
>>92187878Here are the best ones.
>>92187834
>>92187856Sadpanda users continuously assblasted that their tantrums aren't making AI art go away.
>>92187900This right here is the definition of "AI slop" and is no different from people who take 1000000 screenshots of their character in a game and shit up their steam with them. Generating dozens of samey images, instead of making 100+ unique gens is the least creative, most unimaginative thing one can do with this tech.
>>92187947People actually use it for anything other than doujins? what the fuck
>>92187947I'd rather it had AI category though, right now 99.9% AI uploads are complete jank that people would be too ashamed to even post here anonymously
>>92186143self.save_state: bool = False # OPTIONAL, is the intended way to save a training state to use for continuing training, False to ignore
>>92186308Anything v3, v4 and NAI all got flagged as viruses for me.
>>92187969That's still Maine, Anon.
>>92187776many thanks anon
what does the Just resize (latent upscale) option in img2img actually do? I can't tell any difference.
>>92188233uh oh
>>92188279isn't that only for inpainting?
i like cute slightly androgynous android girls.
>>92188294I wouldn't know, I've never understood what it's supposed to be used for.
Am I doing this right?
>download models>don't know which to use
>>92188407>>don't know which to useStart with deliberate_v2 . Very versitile.
does anybody made a model or lora for gore? cus currently all checkpoints i've tried can only generate bad looking skeletons or some weird ropes/sausages when prompted something like "human corpse hanged on a tree, intestines, blood, organs, rotten flesh"
>>92188468try adding horror, dripping, pestilence
>>92188407do a set of gens with the same prompt and seed on each of them, and see what you like the look of best.
>>92188393Reminds me a bit of Mercy (overwatch)
>>92187765I like right they're both good for their style
How do I do step 2?I was under the impression it did that automatically since I added -gitpull so it must be something different that what I think it's talking about
>>92188558I find it funny how well anything holds even now
>>92188528>horror, dripping, pestilence
>>92188558Now I just need to decide on a prompt ...
>>92188528Little better, but still extremely cartoony, and still mostly same weird skeletons.You can't really fix it with prompting, because original dataset didn't had any irl guro or even realistic art. So it can only be fixed by training.
>>92188751Maybe try https://civitai.com/models/18399/body-horror no idea how good it is though
>>92188751Try "art by Suehiro Maruo Shintaro Kago Trevor Brown Toshio Saeki"It's what ChatGPT gave me for guro artists
>>92188689Git pull?
>>92188857how did you manage to do that without it lecturing you about violence le bad
>>92187604I DMed hiroshimoot and here's his response:こんにちは、私の名前は広島長崎です、私のloRAを使って4chanを糞にしてください。以下、リンクです。https://civitai.com/models/21510 ありがとうございます。
>>92188893Oh, it did
>>92185452I have done it, I have trained the Kokonoe
>>92188922lmao, every timeone time it lectured me about using the microwave to kill bacteria, god fucking damn
>>92188940You trained an anime girl that looks like any other anime girl goddamn bro why even waste the compute training just download anything and a vae and use pink hair and cat ears tag same shit
>>92188965nigger
>>92188965anon i'm using the anyhentai modelthat's why
openpose editor vs 3d openpose editorwhich one?
>>92188965based
>>92188965peepee pooopooo
midjourney is way better, close the general
>>92189054https://civitai.com/models/86/openjourney
>>92186089catbox and how you reach this result? I liked it so much.. good work
>>92189054yes, but you can't use it on your desktop (or train it or use your own models or pretty much anything else)also this>>92189066
>>92188906先見の明、アイコン、プレイボーイ、大富豪、インスピレーション、プロFGOプレイヤー、2ちゃんねるオーナー(元)、4ちゃんねる現オーナー、プロボクサー、自己啓発グル、ベストセラー作家、慈善家、発明家、天才、アーティスト、モデル、俳優、歌手、ダンサー、シェフ、飛行士、医者、弁護士、エンジニア、教師、パイロット、スパイ、マジシャン、ゲーマー。 ブロガー、ポッドキャスター、起業家、スポーツ選手、科学者、探検家、写真家、映像作家、デザイナー、ライター、コメディアン、政治家、歴史家、言語学者、心理学者、哲学者、考古学者、生物学者、化学者、物理学者、数学者、天文学者、ジャーナリスト、翻訳家、編集者、評論家、司書、キュレーター、音楽家、画家、彫刻家、詩人
>>92188886what about git pull?
dead threadmove>>92189113>>92189113>>92189113
>>92189276it's not even full, dumbass
>>92189295it's past bump limit and on page 8, it won't be up much longer either way
>>92189312>page 8>on /g/literally hours left
>>92185594is this a lora or a mix like exp mix?
