Previously: >>8375246The cartoony side: >>>/aco/csdgNo DALL-E generated stuff.No futa.Report trolls.>GuidesGeneral: github.com/AUTOMATIC1111/stable-diffusion-webui/wiki | rentry.org/sdg-link | rentry.org/rentrysdHires Fix: rentry.org/hiresfixjan23Inpainting/Outpainting: rentry.org/drfar | rentry.org/inpainting-guide-SDAnimation: rentry.org/AnimAnonCreate LoRAs, TIs: rentry.org/59xed3 | rentry.org/simplified-embed-trainingList of models/LoRAs:rentry.org/acorepo>Auto1111 extensionsControlNet: github.com/Mikubill/sd-webui-controlnet | rentry.org/dummycontrolnetRegional Prompter: github.com/hako-mikan/sd-webui-regional-prompterTag autocomplete: github.com/DominikDoom/a1111-sd-webui-tagcomplete>Related>>>/b/degen>>>/h/hdg>>>/e/edg>>>/d/ddg>>>/g/sdg>>>/trash/sdg>>>/vg/aids>>>/vt/vtai>>>/y/ydg
>>8397649
End of archive dump
is there an archive for these threads?
>>8397903Did you make this?
>>8397903Forced deepthroat?
>>8397886BBC slave tifa is the natural orderCan you gen her getting glowing large pink materia beads shoved up her asshole and glowing pink eyes?Some hypnosis/charm kink stuff
>>8397870>>8397873they look pretty good. Could i get the prompts?
>>8397875Love this one, very classical Hollywood.>>8397905https://desuarchive.org/aco/search/subject/%2Fasdg%2F/
>>8398037Hopefully the anon is still around that made theseWere saved a month ago
>>8398177
>>8398179
If i want to train for a character should i include pics of their body as well as their face in the data set?
>>8398217For body accuracy yes, but if youre like countless drooling retards on twitter you wont care about body canonical appearances and you'll end up giving starfire bimbo proportions
>>8397932Looks like a combination of Kirsten Dunst and Eva Green
>>8398535I think it's inspired by a really old bondage pic I can't find right nowOr maybe just coincidenceWould be funny to hear from the genner
>>8397932oh hey I made that, my gens are img2img of a 3d render, that one I based on yennefer from the witcher using ip-adapter
>>8398634Thanks, now I see the resemblance hehWhich is the right ip adapter model to use with XL? ip-adapter_sdxl_vit-h or without vit-h? Or something else?I only tried one of these once it it blows through VRAM
>>8398651Replying to myself: vit-h actually works, it uses the smaller CLIP modelSweet
>>8398634>>8398636>>8398638i like these, are they all from a render? i'd assume so because the hand holding the cup isn't fucked up
straight jorkin it at the jedi temple with my lightsaber cock
>>8396728Can you please make a futa one? I know not everyone is into this and i won't spam this thread with requests but it's very rare to see Emma being dominated by female creatures in general.>>8397863Also can you "replace" the guy with woman on similar like left image?
>hop in dude
>>8398705no, go to /d/ for that troon shit
>>8398634>>8398636>>8398638thumbnails made me hopeful that was candid Greta
Question. When you are using inpaint with a mask, and you choose "inpaint not masked", does SD see the masked part of the image? If I take a random assortment of pixels and use those as a mask, would this work as a way to ensure consistent colors even at high denoising?
>>8399379"Inpaint not masked" just inverts the maskEither way, it sees the pixels under the mask because you set "Masked content: original"
new spiderman looking p good
I've been having trouble with outpainting and getting "seams" and I can't get it right.I've been maxing out the blur settings, using Soft Inpainting, and used the Mosaic outpaint add-on as well. 64 mask blur, 256 only masked padding. For Soft Inpaint: 4 Schedule bias, 0.5 preservation strength, 2 transition contrast boost, 0 mask influence, 0.5 difference threshold, 2 difference contrast.How do I make it more seamless?
>>8399433She really does look good being choked
10 internet points to anyone who makes Misty or May from Pokemon, but as real (18 year old) women.
>>8399420ok let me rephrase, is the model aware of what pixels it's not changing when it's doing the latent space thing?
>>8399825It's your lucky day.
>>8399825https://litter.catbox.moe/m74ous.png there's never ending slop of this request type everywhere btw
>>8399854The pixels outside the mask, which are to be unchanged, are affecting the gen as long as denoising strength is <1.0A lower CFG should also help match the color/light in theory since it lowers the effect of the prompt
>>8400333Nice how the oily hair was resolved.
>>8397903Only realistic looking one so far. Did you IP-adapter the girl or the style? How do you get the subdued tones? Most gens are oversaturated as shit.
>>8399074Are there any particular prompt phrases you’re using to get such a photorealistic effect? I’ve used epicrealism as a refiner and it is great for realistic people but I haven’t figured out how to generate a photographic effect like in that image
Any recommendations for a pseudo realisitc model?
>>84006641/3 serendipity, 1/3 bemypony30, 1/3 Zonkey42
>>8400427I'd put some voters in her ballot box.
>>8400492its mostly coming from my custom loras. using all the prompts that pony knows still wont get the effect without the loras.Emmahttps://files.catbox.moe/gqi7er.safetensorsEmiliahttps://files.catbox.moe/t7izuv.safetensors
>>8397883god even with AI she's absolutely horrid
Not slutty enough, must go further
>>8400935i like that you have no problem posting your loras. you remind me of me.
I'm possibly liking inpainting too much
>>8400935thanks for these! ewe appreciate you sharing your hard work. Ive seen in some instances where the lora filename has to be exactly as it was originally written otherwise it doesnt fully work as intended. Can we get the original exact filenames too?
>>8400935thank you for the advice
>>8397863can it improve this composite edit pic i did?the main thing that could use improving is the top of the stocking the area of the shirt where her chest is exposed. if you change the resolution please keep it to at least 1500 wide by 1000 tall. for reference you can find the originals at:https://danbooru.donmai.us/post/show/7639962https://danbooru.donmai.us/post/show/7639963https://konachan.com/post/show/378578please make sure that the hands remain uncovered and visible. if the AI can't do that then don't bother posting the pics.
leonardo ai but still pretty good
>>8401960fuck off
>>8401960https://litter.catbox.moe/donpk9.png
Finally started playing with sdxl stuff. used the zonkey pony realism one. I will have to test with it more. how do i get the controlnet to work for sdxl is there a specific sdxl controlnet?
>>8402145Mayb tis can be helpful? https://old.reddit.com/r/StableDiffusion/comments/1dxh7dz/controlnet_allinone_controlnet_for_image/
>>8402479>>8402484For celebs you need LoRAsThe faces are far off the mark
Henlo. First time posting here. Took a while to find a thread that suits my gens, I think this might be it.
>>8401552The file can be renamed freely, it doesnt affect its use as long as the lora callout in the prompt matches the file name.
Fixing nipples is just too damn fun
>>8402090thank you kindly!
>>8402007if requests are frowned upon then that should probably be stated in the first post.
>>8400271There was a gen with woman wearing a huge strapon in this style, anyone have it? Asking for a friend
>>8402693the fucking blanket people... nightmare fuel
i redeemed code for this
>>8402507thanks for clarifications
lol i type big boobies
>>8403031You can shape the body any way you want: https://danbooru.donmai.us/wiki_pages/tag_group%3Abody_parts
before
>>8403051after
>>8397885How can I make those?Is there a way to do it locally?
>>8403088Read the OP. Its all in there.
>>8403120Unfortunately I'm too retarded to do what's in there.I was hoping there was a magical auto installer or something like that.
>>8403098pristine
>>8403121I can't be arsed to do all that stuff either (nor do I have a capable PC). I use perchance.org even though it's nowhere up to speed as sd or dalle but I still have a lot of fun playing around with itwww.perchance.org/text-to-imagewww.perchance.org/proimagegen
>>8403121The very first link in OP is for AUTOMATIC1111, which is wiki where the very first links on that page tells you how to download and run the one-click auto installer depending on your video card.Read the OP. Its all in there.
>>8403121Even the second link in OP, the rentry one, the very first link goes to fooocus which is even more beginner friendly and has an even easier one click installer since then you wont need to setup python since its embedded. You're not retarded. You're just massively lazy which is magnitudes worse.
>>8403121download stability matrix, it will autoinstall various UIs and work as a package manager for add-ons from civitai, it even has a built in inference page so you can generate without installing anything else, it's just more basic than other software
>>8403150>>8403172>>8403174I finally did it but all my prompts look like garbage and abominations.
>>8403195Unfortunately we cant spoonfeed you creativity. Just go on civit and open a picture you like and study the prompts and generation settings they used. You know, READ.
>>8403585Fuck yes, I'd love those for a titfuck
>>8402808I can make one, any suggestions?
>>8403121>I'm too lazy to read just do it for me and build a machine where I press a button and it knows what I want.
>>8403816
>>8403928Top notch baking, as always.
>>8403932Lemme inLET ME IIIIIIIIIIIN
>>8403930holy SHIT
spoonfeed a newfag here. How do you guys get such crisp real looking iamges? whenever I generate they have fucked up faces 100% of the time and the image has fucked up parts all over. im using automatic1111 if that matters. and autismmix. what am i doing wrong?
>>8404482adetailer helps with deformities but it's mainly persistencegenerate lots of pictures with your prompt until you get a seed that you like with minimal body horror, make variations of that seed and then run the best one through img2img until it looks good
>>8404482What size are you genning? Starting small and upscaling can help
>>8404512im genning at 1024, I should be generating at that resolution for sdxl models, unless im mistaken?>>8404499ive been loking at adetailer. does it work for multiple subjects at once, or just 1girl scenes?
I want to make a custom lora to insert a certain person's face in my images.How can I do that?I already have tons of images of that person.
>>8404482every single one of us started by seeing fucked up faces and body horror (or that one asuka test prompt when NAI first dropped)… just takes some time and testing to figure out what works best, with some tweaking for different models etc. also keep in mind that SD 1.5 and SDXL models have somewhat different prompting than pony models, and autismmix is definitely a pony model. I use pony realism but that requiressome extra attention to detail and maybe a couple of loras, so starting with drawn styles in autismmix is a great way to learn. go forth and create!
>>8404523>>8404482There’s so much to know it can’t be spoonfed all at once. I’m still fairly new but am way more capable now than I was even a week ago. You’ll learn the basics and look at images and see what’s possible and find out what they did then learn how to do it yourself. You’ll start rough then get better as you go. I’ve been using pony realism and I find that using very simple prompts at first is a good way to understand fine tuning the generation settings (1girl, wide hips, large breasts, nude, posing/kneeling/standing, white background).Adetailers will automatically detect all the elements in an image that they’re designed to find. Face adetailers will detect and inpaint all faces it detects.I believe good images are the result of a series of processes, some automated like adetailers and upscalers, and some manual like inpainting.
>>8404740adetailer is literally just automated inpainting btwbetter to inpaint by hand for a while so you learn which settings you need for your adetailer
>>8404482short answer: copy prompts from civitailong answer: there are only so many pixels in a face and the AI is picking from a lot of possible solutions for your overall prompt. to some extent, you can avoid undesirable results through prompts - e.g. if you have depth of field in your prompt, you probably won't get bokeh, or through negative prompts, like putting totem pole or amputee in your negative prompt will avoid head stacks and half limbs. However, if you are generating at 576x768, if you prompt for a portrait, your face will probably be fine, but if you prompt for a full body, the face doesn't have enough pixels for the AI to make one that passes the smell test. Prompt weights can also be a factor; increasing a weight too high can cause adverse effects. (Gigantic breasts:1.5) can give you an image entirely of tits. Inpainting the face with the "masked only" setting will expand the masked part to the full size of the image, and then modify that. This is particularly useful for faces because of the aforementioned pixel thing. You can do this manually in inpaint, or automatically with adetailer.Post some of your prompts (and model) and I'll see if there's anything obvious. Negative prompts are particularly important with some models.
>>8404883>>8404482The other thing is that image dimensions have a LOT to do with your final image, more so than your model, even. If you're genning at 1024x1024 in a 1.5 model, you're going to get body horror, because it's trained on smaller images, and instead of expanding them, it will just fill up the extra space with things from the prompt (including things like "standing" or "tits"). >>8404499fwiw i almost never seed hunt
>>8404888excellent point, I found this helpful https://civitai.com/articles/3962/how-to-creating-higher-resolution-landscape-images#heading-5460
i merged zony realism and pony realism checkpoints
i meant zonkey
>>8404917you notice that some of the realistic checkpoints are more resistant to actual sex acts? Been merging a few as well, pic related.
>>8404934not yet. i havent messed with merging XL checkpoints until today. this one doesnt have issues with sex acts at least
>>8404926this does look great, nice work
>>8404934nta but something i noticed in realisticvision is that the models can do sex acts, but the tokens are extremely weak and will be overshadowed by basically any other prompt. I can gen sex acts if that's literally the only thing i prompt but once i start adding style prompts it turns into a generic 1girl, even at 1.5 strength
>>8405171Are you using the danbooru tags for the positions?Order in the prompt matters, move it closer to the startMost pony models need score_9,score_8_up at the start, which will help with compositionWhat kind of style prompts are you trying? They might just not be supported by the mergeIf in doubt, stay with danbooru tags
>>8405190i'm talking about a photoreal 1.5 model
>>8395204>>8395354Wow, this Goddess model is really great. What kind of language prompts it follows? I couldn't find the pattern in sample pics.
Whenever I try to generate prompts from a text file, it doesn’t seem to register the sampler I specify. That is to say the image doesn’t generate correctly and the sampler is not listed in the png info at all. Does anyone have a template that they’ve had success with for generating using a text file, or at least know how to specify the sampler?
>>8400271Fuuuck! More BWC-CBT pls!!
>>8405248My bad, realisticvision is obviously not ponyYou are not going to get good custom hardcore porn images from non-pony modelsThere is simply not enough training in those for it to have learned those promptable conceptsObviously there are porn images even in the base model training setsSo what you see sometimes is how it's trying to make a sort of remix of those imagesBut you can't really customize details like poses, sex organs, number of partners etc.Just use pony for that, it does it all and more>>8405269Pony models
Is this thread only for "realistic" generations?
>>8406086No, but anime-stye generations are more the métier of /e/, /h/ or /d/.
Gwen mask lora, trigger "gwen mask", adding "mouth mask" can help, as well as putting "mouth, ears" in negative.https://litter.catbox.moe/zc6fag.safetensors
>>8406222You could not have chosen a more disgusting content to gen to advertise your Lora with. Piss should be banned to /b/
>>8406249objective achieved, pissing off insecure kids
>>8406222Is this the beta already posted or did you finish baking a new version?
>>8406086That can go to the cartoon thread >>>/aco/csdg >>8399151Semi-realistic renderings are kind of a toss upThis thread here is meant for photo or near-photo realism
>>8406291Pony version
>>8406222Thanks, prompts correctly in zonkey 5.0 every time as long as I include a colorI noticed it hijacks arms above head and makes her touch her head insteadI'm working around that by adding the mask with adetailer after the fact, which also gives me more freedom with colors and materials without resorting to regional prompter
https://www.rollingstone.com/politics/politics-news/aoc-deepfake-porn-bill-senate-1235067061/So... rip these threads?
>>8406498Most gens would never pass as realThis should concern mostly nudifying of real imagesWhich already got a guy in jail in bongland cause the girl was minor
>>8406508>Most gens would never pass as realI don't know if this part is true. The bill specifies that a "reasonable person" could think the images are real and we know how bad "reasonable" boomers are at detecting even really shit AI images
>>8406519It won't be up to boomers it'll be up to Mark, Ken, and Ahmed in the office who have to sort this shit
>>8406519It's actually even worse. That law says that even labelling it as fake is not enoughI wonder why they include the other conditions then. Typical law maker retardation
>>8406498Always post celeb pics without labelsLaw fags can figure out the resthttps://en.wikipedia.org/wiki/All_persons_fictitious_disclaimer
>>8406541I think it's silly to assume this would save you from the wrath of any feminists
>>8406547But they would go after people that tweet those pics at them firstAlso the law is worded for revenge porn/regular people, not public persons/celeb casesAnd they would shut down civitai first
>>8406498>Congress finds that:>(1) Digital forgeries, often called deepfakes, are>synthetic images and videos that look realisticEasy fix: From now on, gen all celebs with giant milk tits, gaping holes, giantess/minigirls, futa, horns and wingsAlways happy to comply with US Laws!Let them explain in court why a reasonable person would believe that extreme prolapse was real>>8406547>Caring about feministsLol USA
>>8406584Every AI image includes a UFO in the background, hence not real. If it is declared real, then UFOs are real. BOOM
>>8406586That's the big hole in this law. They should have worried more about logic than spending multiple paragraphs on codifying victims feelingsThey state that marking it as fake is not enough to be safe (what about Free Speech though...)But at the same time it has to be a realistic imageSo drawings are completely okayAnd no matter how photographic it is, if the content is absurd enough it won't be realisticAlso dead celebs are okay since the person has to be alive to sue
>>8406644I don't see any carve out of absurdist. Only "realistic" and "lifelike". So if someone generated Leo DiCaprio naked that would be illegal, adding an alien to the image isn't going to save it.
>>8406655"Your honor, this digital forgery depicting my client is causing harm to his reputation and so we seek damages""How so?""It shows my client having intercourse with sexy aliens from outer space""And?""It's of photographic quality and so people could think it's real!""..."Just make sure that the test case that makes it all the way to the Supreme Court involves proper futa so we can all keep laughing at your joke of a country
>>8406876really nice style, share a prompt?
>>8407018Using this model: https://civitai.com/models/597602/lusty-weekend-at-zavysI basically copied the sample prompts.>close up, Stevan Dohanos style painting, Stevan Dohanos painting of a jedi woman, wearing a sexy robe, with a lightsaber, see through clothes, 1960 style, (((by norman rockwell, Stevan Dohanos))), in style of Saturday evening post, classic pinup art posing, rough colored oil painting style, vivid color, thigh gap, <lora:add-detail-xl:0.3> <lora:NsfwPovAllInOneLoraSdxl:0.3>
>>8407049thanks anon, that Stevan Dohanos was a fucking gangster, appreciate it!
>>8406995you're underplaying how much harm someone can do to someone with this tech
I need some more AOC’s to add to my collection so that they give me the noose when they catch me
Can I make a request from someone who's better than me with Pony? I'd love some kind of public use bondage, preferably with creampie, in classical painting styles.
what's /asdg/ recommended models for realistic images?
>>8407460deeznutsXL
>>8407490Why is it so appealing to gen Hermione getting used like a hooker
>>8407490WOW incredible, can you do it with scarjo or emili clarke
Shout-out to any film connoisseurs in this rancid general who recognize her.
>>8408080It's not easy getting a convincing Milena is it
>>8402484>>8402479What's going on with the aspect ratio there
>>8408329Common beginner mistake, mistyping the resolutionFor A1111, I recommend sd-webui-ar_xhox extensionIt lets you pick resolutions just like in Fooocus
>>8408377Looks like dalle shitShow us where the milk comes from
>>8407460Lustify is a good model for porn aesthetics.
>>8407826Yes. Emilia lora is working. Guess ill make a scarjo one from dalle thread archive.
>>8408447nice!! Yes Im also intersted in seeing a good Scarjo one from you! Thanks in advance
>>8407826>>8407593Shut up>>8408080Beautiful
Toatazles not montroonn
>>8408541Maybe a bit over the top
>>8408745hmm forgot to upscale
>>8408080these are amazing, I demand more!
>>8408746>>8408747>>8408748>>8407400>>8397903holy shitI was just starting to think my gens are decent lmao
>>8408753
>>8408808
>>8408821
>>8408850
>>8408808>>8408821>>8408850>>8408875I'm fairly certain Milena Velba like the girls m8.
>>8408939retarded ESL
>>8409005*likes*1 spelling mistake != esl you fucking illiterate cuntswab.
>>8409037more casually nude office girls
>>8409078>>8409076>>8409074more of this please lord
>>8409006stop making spelling mistakes then, esl. you're still retarded for thinking she's a dyke, too.
>>8409104>you're still retarded for thinking she's a dyke, too.>t. has literally never watched a Nadine J video...
How about some brace face blondes with pigtails
>>8407981Is that Texas chainsaw? She looks like the radio DJ
>>8409111dude, SHE HAS CHILDREN
>>8409136You think lesbians can't have kids either by adoption or previous relationships?
>>8409136yeah and most of their vids feature WOMEN sucking lactating/non-lactating tits of other WOMEN and (ocasionally) some pussy play.bi maybe but defo enjoys sucking a milky tit. Why does it matter so much to you?
>>8409140He's afraid his imaginary girlfriend isn't into him
>>8409147>be anon>buy you-beaut, top of the range pc with duck nuts gpu>spend months on github, 4chinz and who the fuck knows where learning to gen>dld gigs of models and loras>get a decent library of Milena because macromastia and lactophilia>start genning pics of Milena>she says, "Yeah... nah."
>>8409138>is coffee gud for you?
july has been a heavy month for me generating.so i have heard a rumor about Ai tools getting released in a big way after november when the election in US is over. I believe it.
>>8409135>Is thatbingo!
>>8409326Requesting tight Japanese sluts getting blacked
>>8409329
>>8409408We can go further
>>8409162She's actually running a retail business in Prague these days
>>8409312I'd love to see her little bottom in hemmed, loose, but deep riding denim shorts. Sort of a holy grail for me. The kind girls wear at the beach with a thong, but no frills.
>>8409449So does that go in one hole or both
