Previously, on /asdg/ >>8578561The cartoony side: >>>/aco/csdgNo DALL-E, Meta AI or hailuo AI generated stuff.No futa.>>>/aco/futasdgReport 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>>>/trash/slop>>>/vg/aids>>>/vt/vtai>>>/y/ydg
>>8591459nice proportions cuh
And that was the random archival dump.
>>8591459I like your style my dude !
>>8591459that's really beautiful
>>8591692
some DS3 waifu
>>8591747
>>8591749
>>8592107
>>8592110
>>8592111
Rentry link is down. Anyone have another one?
>>8591463
>>8592306No, and that means we´re fucked. It was almost all /csdg/ stuff.
>>8592618more of this model?
>>8592427Nice, is that a Lora or just prompting? If so what model
SD babby here. WebUI user. I've done a little bit with LoRAs, inpainting, and Roop, and that's it. Basically a scriptkiddy.You know how FaceApp when it still existed was able to gender swap, or change age, or change race of a face and still have it stay recognizably that specific face? For instance that one guy who kept trolling people on twitter with "and this is how you would look if you were black or chinese!"?I want to get stable diffusion to do that.Start with a face, alter the face based on a prompt, but keep it recognizably that same face instead of generating a brand new one.
>>8593030That can be done with just inpaintingAdding some controlnet model can take you farther
>>8591747>>8591749>>8591750good shit, love the dancer. how did you animate those?
>>8593207inpainting would just give me a random face that fits the prompt, no?
>>8591458requesting Brittany Venti fakes please bros
>>8592867danni ashe pdxl at .90 and cammy sf6 at .20 both are on civitai
guys, we can back on fun with kamala now kek>>8593350she's available on civitai. just create an account and try some gens
Someone genned a few weeks/months ago picrel using ciciai (a dalle front that was uncensored at some point), is there a way to get the same style locally?- something that could have been posted in early social media.- has that old amateur "university dare" feel?
>>8593300No, if the denoising strength is low enough and you keep the original content it will mimic what is already there
>>8593673good erotic hip swaying
>>8593527I don't think anything local can replicate dalle "early nsfw social media" style yet
>>8593439>try some gensmind explaining this to me like im some low iq retard?
>>8593942
>>8593972>>8591458>No DALL-E, Meta AI or hailuo AI generated stuff.Take your crap to its containment thread.
>>8594022quit your bitching with your ugly ass face posts you constantly post
>>8594022annoying dumb bitch faggot i hope you kill yourself so you stop self moderating 4chan threads like the annoying dumb fucking moron that you are. literally slice your throat open and die you stupid bitch
>>8593300Try ip adapter.
>>8593673how ? The boobs animation is surreal
>I was fond of our maid Clea and her new baby, but after she accidentally broke a treasured vase my mother would have fired her for, I told her I'd accept the blame in return for taking her son's breakfast every morning for the next three months. She let it go on for eight.
>>8594520
>>8591747>>8591749>>8591750Damn, these are great, especially the lactation one.Not to r fag the thread, but any chance you're into the DS3 firekeeper too?
>>8594525>>8594527>>8594528>>>/aco/csdg
>>8594517kling
>>8594539
hey hi hello
>>8591775nice, prompt/model?
>>8595414her big toes are the wrong way lol
Posted this in another similar thread but I'll ask it here, too:Why do you guys insist on posting this stuff on the CARTOON porn board? Like, why aren't you posting this on >>>/s/ ? It's clearly not cartoons, it feels like it should be there instead, right? Real question.I'm not even arguing that it isn't good or that it's bad because it's AI, I'm asking why you're posting realistic generated stuff on the adult cartoon board.
>>8595808For historic reasonsThe thread was made by /tv/ refugees making celeb gensI don't see any AI thread on /s/, are they friendly?
>>8595826Apparently /s/ is not the right place >>8595834I think /hr/ might be fitting, but they require images to have a resolution higher than XL, so everything posted would need upscale/hires fix
>>8595826"Historic Reasons" isn't really a good reason, though. Also, /s/ is a board that's always had one foot in the grave. To say they're friendly is like saying /po/ is friendly. They exist to post porn and barely anything else, when they're actually posting.>>8595840I wouldn't assume /s/ couldn't be the right place, given the testimony of a single person getting really angry at nothing for no reason. There's no reason to assume they wont be welcoming without anyone even trying to make a thread there.
>>8595849Historic reason is a perfect reason >>8595794It was probably among the 6 boards he tried back thenComplaining about this being here feels autisticIdeally, 4chan should finally introduce new boards again and a dedicated AI board should be among them
>>8595840>>8595865Why do you keep linking to other threads? Why not formulate your own responses? Also, I'd love to hear why you think "historic reason" is a perfect reason for it. Historically a lot of threads existed for years on specific boards that eventually got removed from those boards when the mods realized they weren't actually appropriate.
>>8595898Because someone already tried what you are proposingHe ended up on /aco/, so why repeat history?I'm linking the other thread cause there's more activity there and they are making good points
>>8595808because ai is not reality. you will never meet, the people you see here. ai is like 4D or something
>>8596032>you will never meetYou'll never meet the people you see in porn or in the movies either, Anon. Real people in porn may as well not be real. The issue is more so that they aren't cartoons. But I'm just going to drop this issue, it's clearly it's making some people uncomfortable.
>>8595898https://www.4chan.org/rulesThis is the reason: /s/ classifies AI-generated images as fakes. Also, if 3DCG is allowed on /aco/, then 2DCG is on topic too.
>>8596083*ahem*
>>8596387Also no hardcore and fetish stuff allowed, so /s/ is totally useless for usThe guy is most likely a mod btwEveryone that called him out hard got their posts deleted or a warning
love gooning to Mommy Elvira
>>8596083wrong. a friend has met many actresses in the past, and he's not even rich. good luck meeting ai women from here lol
>>8596587Tits are too small
>>8596387Fuck you JEW CHAN!!!
I'm seeing some pretty good ai video porn stuff now. What are people doing to generate that stuff? or is it still in the early stages where you need a supercomputer?
>>8596938>>8594000
so i downloaded stable diffusion.how do i actually make it generate porn and not look like shit?
>>8597548gotta use prompts that work with the model. for instance if you use pony realism, you can use booru tags as prompts.and make sure you adjust the parameters to whatever the model works best with, like the render size, sampler, cfg scale, etc.
>>8597555whats the best sampling method
>>8597578depends on which model you go for, but most of the time DPM++ 2M SDE Karras is the one alot of em work well with i think.could be wrong though since i'm not up to date on all the sd happenings.
>>8597590the stable diffusion i downloaded doesn't have that one.is the model the same as the lora
>>8597578you're going to wind up doing your own experimentation anyway based on lora's and the strength of tokens, you're going to have find a balance between clip skip,steps,CFG, and sampler.
>>8595808/s/ bans AI and fakes
>>8597593i'm unaware of a pony lora, but in general lora's are different from checkpoints/models. they serve as assistants to generating things like styles/characters.also i'm unsure of how you don't have that sample method, but i'd suggest trying to update, or use sdforge which is a bit better imo.
>>8597593(Base) model AKA checkpoints are what determines the resolutions that you can work withThe common base (vanilla) models in use are SD 1.5 and SDXLThey do photos and drawings but they suck at nudity and pornSo people made finetunes of these models trained with porn images for instanceThis is bigLust, bigASP, Anteros, lustify etc. and especially pony (which is a bit special again since it's trained on cartoons and then gets merged with photo models again)Each checkpoint is 2GB (SD1.5) or 6.5GB (XL) to download and you can only use one, so people invented LoRAsThese are much smaller and each is trained with a specific celeb, clothing style or art styleLoRAs can be combined to a certain extend and are compatible across models derived from the same base
>>8597851what am I looking at?
>>8597894One of the most realistic images posted in the past few threads.
>>8597851yeah, i'm gonna need proof that this is ai and not some random amateur.
>>8597968The eyes look wrong, especially her left (our right). I also feel like there should be some red eye. The dimensions of the image look like SD, as does the filename. Her belly button seems kind of high up too.
i'll be 1girling on the "idealized" side for now.
>>8597590DPM++ 2M SDE gives a more natural, less smooth skin tone, but gets flaky with not enough steps. Euler Ancestral will prettify things, but will give a more artsy, less natural feel. The others are just left overs from earlier experimentation ^^
>>8597948>>8597968Fucked up eyes, arm, knee, ear, pubes7/10 maybe>>8598049People ascribing these kinds of observations to samplers is peak snake oilDo you perform rain dances to affect the weather?
>>8598045May I have more like this but with pubes?
>>8591458FLUX.1-schnellShellyLong LoRA
>>8598221bit of a gamble, pubes usually cancel the panties. but here's one that worked
>>8598378very nice
>>8598351No panties is fine with me, bring on the forest.
>>8598879Britney Spears vibes for some reason
>>8595808Personally I think the site needs a dedicated AI section unto itself, but what do I know.
>>8591458is there a reupload or a relink to the acorepo in the OP thread?anytime I click on it, it just takes me to a 404
>>8599901No. But it was all /csdg/ material.
Hello, can someone offer help on what to prompt to get this pose in Pony Realism? I can't get clothing to cover just the breasts. Almost always it covers the bottom and completely exposes the breasts. I want the clothing to be over the breasts and the hands to be over the clothing.
>>8600869Easy way out: just controlnet with an image.Prompting: I´d go with revealing clothes, open shirt, covered breasts, hands over breasts, navel, midriff.
>>8600869>cleavage,open shirt, bare shoulders, hands on breastsGets you pretty close (maybe add nipples to negatives)If you try enough seeds, maybe the clothes end up under the hands sometimesIf you don't like randomly trying, go for controlnet and/or inpaint a bit (sketching the shirt color around the hands might fix it in img2img)
>>8601025Incredible. Model/Lora/Technique?
Are there any discord servers or forums for general hentai / realistic AI porn creation?I tried to search, but most of top discords are dogshit falling under some company/org that allows only content generated from their platform, and there is not really any useful discussion.
>>8593527dawg nothing is gonna be as good as what u just posted for that kind of vibe but i also like making that kind of stuff but it is difficult
Can anyone make a post like this of my hot ex? Is this legal?Kik Jason1446
>>8602570>>>/r/
>>8603050kek should send them to /b/. they faggots welcome beggers like the cum gargling faggots that they are
>>8603072You insult /b/ and a wild pedo poster appearsLike clockwork
hell yeah
>>8604204
>>8603651is this inspired by Under the Skin?
Requesting;-Taking your (human) pet for a walk gens-a van or bus with the words 'Simon's Slavery Services' on the side , in a suburban setting , with the handlers guiding one or two naked slave boys into the vehicle.
Ai Nimoy is now monitoring this thread. You guys better behave.
>>8604535
>>8603651Strangely hot.
>>8602559yeah I tried locally but I'm unable to reproduce that erotic + early spicy social media student dare stuff.
>>8605253I´d try with tokens like: vintage 90s, (polaroid:1.3) photo, grainy photograph, 1990s photo with film grain, 90s photo with vignetting, retro, 90s photo, vintage photo, 90s home video, nostalgic 90s photo, taken with kodak disposable camera
thank you for the abbey x BWC
Is stablediffusion still the way to go to get into making ai art?I tried it over a year ago and had some fun but it would take minutes to generate 500x500 images and it would crash if I tried higher.I'm upgrading my pc soon. How much vram is expected for genning?
>>8606447For uncensored local? Yes, just pick a webUI (forge, Automatic 1111 or ComfyUI if you are node nut). There are other options. From China. Heavily shilled.VRAM: the more the better. For the current models (Flux, SD3.5) you´ll need a lot of juice. If upgrading, attempt 12+Gb for some future proofing.
>>8606323>>8606325>>8606326Nice
>>8606448I see thank you so much!
>when you see a guy, adding a lora of his real face on civitai
>>8606326Can you do these beauties as elves?
>>8604535More Heroic Sar Trek (original ) AI plox.
>>8604226>>8604212>>8604207Very nice stuff anon
>>8606664fucking hot
>>8597848>LoRAs can be combined to a certain extend and are compatible across models derived from the same baseI created an original (homely) character by combining several SD1.5 models (generic female character, shortstack, big round nose, plump lips) at specific strengths, genning a number of images then using them to train a new model in both Pony and Flux. I chose the wrong epoch of Pony model and need to look at the others now but the Flux one worked out pretty well.
>>8606762God fucking damn it wrong pic
>>8606768The reason to bother with training character LoRAs is to get subsequent consistent images in other situations.
>>8606664I tried unsuccessfully to put the BoF Elsa Lanchester hairdo on a character recently, did you do it entirely from a prompt or was a LoRA involved?
any Dafne Keen?
>>8606803you can create a lora of any human on civitai
>>8606539He just wants to fuck all your waifus
>>8606762kek at the wrong picYes you can in fact use multiple character/person LoRAs at once to create new peopleWhat I meant by that comment was the case of using a person LoRA with a pose LoRA for exampleGood job on the training with gens, it's fun to be creative with thisThere is no right or wrong way of doing things
>>8606984I made my first SD1.5 LoRA by using command line tools inside the Kohra (Koyra?) folder to fuse the original models, learning they had to have compatible "resolution" to be able to do that and wasting a lot of time I could have spent instead having civitai train off of 10 images. I would call that doing it wrong but it worked and I learned something. Downside is it doesn't look right on other sites using the same checkpoints whereas the Flux version trained from images looks right everywhere.
>>8607142In the end img2video doesn't care which one you use as source, it's still a crapshoot
>>8602777Is this that HI\/E dude's OC with the butter knife for a nose?
>>8607156>>8607154>>8607152These are amazing!
Hey yall, I am pretty new and learning to use Stable Diffusion and comfyui. Learned enough about steps, cfg, sampler, scheduler, etc that I am making progress making better looking pictures. Where I am stuck now is if you are happy with 90% of an image but 10% of the details where its converging is no longer improving.For example, I have a picture where the face has some harsh pixels so I tried putting it through img2img with inpainting but I am basically getting the same face again. What do I have to change to make it look better on the second pass ?
>>8607258whew finally got something decent. what i did was drop cfg to 0 and max out denoise at 1.00. then I just went through seeds until i found something I liked. also used a sampler that supported linear quadratic scheduler so the new body part would be painted on.what I had trouble with was lowering denoise didn't really blend anything but keeps throwing back what you already have. I also changed the mask to just one facial feature, doing the eyes, nose, and mouth all at once wasnt coming out good.
>>8607316oh i feel dumb. i didn't realize there are models/checkpoints specifically for inpainting. (they are good at generating parts of images) so i was also struggling because i was doing things the hard way.
>>8607258Use inpainting on the parts you don't like (img2img applied to only a part)If you get the same thing over and over you fixed the seed or denoising strength is too low (it sees what is already there and doesn't diverge enough)As an alternative to manual inpainting, you can use a Variation SeedWhen you found a gen you like lock the seed, turn on a low percentage Variation Strength with a random Variation SeedAll the next gens will differ slightly and only change little details, which eventually fixes glitches like bad fingers etc.
>>8607196thanks
>>8607057>>8607058CUTE!
>>8607517Please keep making more.
I gotta bunch
>>8607153Damn, I'd germinate her fungal spores.
>>8607749>>8607745>>8607731Do you take requests? I'd love a lewd naked Norman Rockwell style elf wife! As busty as you can make em.
>>8607839I'll look into make some tomorrow. Kinda hard to control breast size with flux in my experience.
>>8608096Pretty amazing she can ride a surfboard in heels
>>8608083Thanks for the work, I love how they look like they came from a pulpy 80s artist's coffee table book.
>>8605354thanks I'll try it
Some people would do anything for a cabinet spot
>>8608525Which model are you using?
>>8608629Nice pussy, post more hairy chicks.
>>8604622>>8605524>>8605525Which model and what prompts did you used?
>>8591467Great use of dynamic posing.
Can anyone help me with this? Since yesterday Fluxgym can't get past epoch 1, before reaching epoch 2 it finishes and says it has just completed training.I've uninstalled it twice and installed the latest version of python and the problem continues.
Testing big love pony model, seems pretty cool. this pic took 30 seconds (including face adetailer) with rtx 3060. It only requires 8 samples.
>>8609939Are the arms always fucked up at such a low step count or did you gen unlucky here?
>>8610003lol, I didn't even notice that pic had fucked up arms. I just took some random picture from the output folder without looking at it closely.But yeah, I think it was a "unlucky" gen. That being said, I think the biggest issue with this model are the faces. SDXL/pony faces kinda suck.
Zatanna
>>8595547Don't a toesist! She's Antipodean.
>>8591680>>8591681>>8591683>>8591688>>8591689such a pity that they look so retard with the MAGA signs on
>>8607057ARI SALL ELDET!!!DAIS VAULT!!
>>8610460niceI end up doing anime characters way more than generic girls or real onesyou left a foot in the corner tho
>>8610493That's the last guy she wore out. Fainted on the floor from fluids loss.(Ask me how I know)
Can I get some NSFW AI story generators recommendations?preferably with lorebook support
Could someone guide me on how I could create animations using AI?What program do I use?The link available in the thread does not work
>>8611281There is a thread in /g/ dedicated to local language models>>8611362reforge has some video plugin installed out of box, but never tried itI think most people that make videos use some free service
>>8611419but do free services generate nsfw videos?The ones I tried don't allow it
>>8611427your a year or 2 still too early. you wont find anything that works the way you want it to.
Does this still suck on AMD GPUs? I'm considering swapping to NVidea for this and VR purposes.
>>8612827Frankly, IDK, but it seems it still takes some work to use AMD.
>>8612827Works fine on Linux, but probably runs slower than NVidia and uses more VRAM, so you can do less stuff with comparable hardware
This is raw output from InvokeAI with my prompts. I just started learning how to do all this yesterday, so sorry if it isn't very good (or not Western enough for the board).
>>8612827nVidia is better, but you can use ZLUDA to use it on AMD GPUs without resorting to Linux.
>>8612888I'll post some more I made.
>>8612928
>>8612929
>>8612931
>>8612935
>>8612939
>>8612928Dude, what´s with the futa spam. take that to >>>/aco/futasdg
>>8612952Oh, sorry. Didn't see that thread. These are all guys, anyways.
>>8612962Then I recommend you looking into >>>/y/ydg
>>8612964I am both blind and retarded. Thank you.
>>8612888>>8612928>>8612929>>8612931>>8612935>>8612939>>8612940>No futa.
>>8613297didn't he already apologizejust let it be
A while ago I asked about adetailer bogging faces in hires gens, when the low resolution versions looks perfectly fineI found the solution myself and it was quiet simple:For hires you need to increase the "Inpaint only masked padding"It simply sees too little and misjudges the camera angleAlso always choose the lowest denoising strength you can get away with for the same reason
>Waste 3 days trying to create Anna Maria Marconi from Superior Spider-Man in SD1.5>End up accidentally creating a completely different looking OC "X" who turns out great>make good enough gens to train a Flux LoRA model of X>Flux is useless for porn, gen a bunch of Flux "X" to train a Pony model>Pony X model looks like shit, drop in a couple of standard negative embeddings just to clean up errors>Prompt now inexplicably no longer looks like X but somehow is Anna Maria Marconi
>>8613828
>>8613828how looks your flux lora?
>>8608551For the record. JuggernautXL 5.0
Baking.
New thread:>>8614027>>8614027>>8614027
