Putting the Sticky in Realistically Previously: >>8122572the cartoony side: >>8159673Previously >>8092147Any Malu >>8121665 >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>>>/h/hdg>>>/e/edg>>>/d/ddg>>>/g/sdg>>>/trash/sdg>>>/vg/aids>>>/vt/vtai
Thanks to the power of AI, we finally get to see Eva Green's tits!
Would you eat the snow queen's turkish delight?
the devil is in the details
Tilda can be a sexy sticc ayyyy chick
Got some neat heroin chic arty photos with her.
We can also... embiggen her.
Holy shit! David Bowie fucked her!Is... is this Twincest?
>>8167116hehe
la la la la laThis is my stopGot to get offI might go popExcuse me, Excuse meI've got to be directla la laIf I'm off, please correctla la la
Would you stand on her neck?
Trying to find some sticky piccies
Give me more Kat Dennings
This error is udderly ridikulos.
>>8167186sureshe comes in bunchesI got my hunches
some of them have very iffy likenesses tho.yet I like the fusion of illustration and photo real here.
There are some nice nude ones that came out... questionable. This one isn't nude but looks great and looks appropriate.
Here. The younger likeness here is good, but I dunno about showing the full thing. I mainly like the weird guy's face.Obligatory>me on the right
Meanwhile, back in 1954...
>>8167228ah... retro porn... or old actresses in new porn?
Holy shit. I have 1.4 gigs of 22,000 unsorted images.
She looks older here... but got a boob reduction.
Damn. Many of these didn't save the prompts as comments and I can only guess how I genned them. Hell. Some of them I genned on a website while in bed on my mobile. I kept saying I was gonna label them and sort them... but that was thousands of gens ago...
Some of them are Asian versions of American actresses. It's like a guessing game who they are.
I have months of unsorted gens and many are from just lying in bed listening to the radio and making gens on my mobile. Some are more fitting of the cartoon thread. And some... some are questionable to post even on /b/.
I wish I could remember which actresses some of these were supposed to be.
> This is fine.Am I posting too many? Just realizing how many neat ones I got in my old folders. Any suggestions for themes?
>>8167177>>8167179you wrong for this lmao do jane next>>8167243i got up to like 3 gigs across five or six different folders before i just said fuck it and deleted them all. i think i had realistic 1girls, lineups, lora tests, medieval painting style, a WIP folder, and one for the early experiments I did
Can you imagine telling your younger self that some day you could go on your computer and have photoreal AI art generated of your favorite actresses doing the lewdest things?I know we've had fake porn made with old Photoshop, but this is a whole other level. Back in high school and college in the late 90s / early 00s I browsed faker sites and even tried my own with my bootleg old copy of Photoshop, but it was nothing like this.
>>8167276it's not that you're posting tooo o mamanny, I certainly have spammed these threads plenty, but maybe you have something besides 512x768 1girls with shiny skin
>>8167282turn yo CFG down fool
>>8167237Why not both?Time has no meaning for AI.Except for the fact that it seems to think people actually looked like paintings in the 1600s, which is pretty funny to me.
>someone uploaded literally a hundred actress loras for pony>its all generic garbage that looks like shit without reactor and stuffbetter than nothing i guess
>>8168812How generic? At least they are flexible, ie, you can get more than two facial expressions?
>>8168902seems to work with no checkpoint but Everclear (the description suggests it works with other realistic checkpoints, Pony Realism looks nothing like the subject)seems to be pretty inflexible overalllooks like the guy who made this trained on SDXL but no booru tagstraining on any of the realistic checkpoints already seems like a bad idea, so far all of them have lost a lot of the flexibility of Pony
>>8168947Ok. Thanks for the analisis.
slow board
>>8169207prompt?
>>8167007OP did you happen to read my post a week ago? I asked for a Kat pic and you delivered, ty
Opinion on Orthodox Icons?
>>8169961pic related
>>8169961>>8169976they don't belong here
>>8168788based fellow best emma enjoyer
>>8168812one of these days I am gonna get into loras and all that fancy stuff... but for now I am trying to do what I can with bare bones SD 1.5.
>>8169893You're welcome. Sadly most of my Kat pics are unlabeled. Only a few of them have the prompt saved as metadata. I used to make a ton of Kat pics back in the day, plus would save a bunch posed, but I want to post more recent stuff. Here, have a cum drenched Sydney Sweeney as I look for more Kat.
Surprised I didn't do Kat for this prompt. Neat thing about some prompts. You can just swap out actresses by swapping names. No wonder actors don't want AI used in films.
>>8167007there's a neat thread on >>>/gif/27097229 that's making stuff like picrel using stablediffusion and replacer.it's also doing boring stuff like deepnudes, but I figure you guys would probably be more interested in making animations look realistic.
Damn. I have a bunch of photo real pics of Natalie Portman, Brooke Shields, and Jodie Foster that look like they are from... uh... earlier in their careers. Though I don't think even the thread on /b/ allows those. Here is one that is more cartoony for example. Though I guess I should post those in the other /aco/ SD thread, here is an example. I just use a different model and switch secondary prompts, it becomes photoreal... and a bit too questionable for even 4chan.
>>8171208interesting thread
Anyway... more more sticky posting... I find sometimes fluid physics are bad, but sometimes it comes out pretty nice.
Last thread ended with a Sansa/Sophie pic. Was gonna post some more, but realized the thread died. Then I posted a more recent Kat pic.
Rarely I get a decent deepthroat... unlike time I spend with my wife. This is why AI is superior.
Am I going face blind from all this AI or do Sersh and Sophie look like sisters or cousins?
when irish eyes are smiling
aha. found some more kat.
>>8171177Pal, you are using the perchance site, not fooling anybody.
>>8167295>turn yo CFG down foolNo. I don't think I will.
>>8171625I wasnt trying to. I thought Perchance used SD 1.5. Is it an older version? I know it is simple and nerfed, but I am curious to see what I can do with a simple site like that just with prompts and limited inputs. I said I want to ~eventually~ get around to doing a proper local install. Just like I want to finish that novel... and Valve gets around to releasing Half Life Episode 3...
Though... yes... it does often leave much to be desired.
I should get back to Kat posting...
hmmm.. lemme adjust the exposure and hue...
I like how you can adjust... certain things even with simple AI programs.Nice.
Hmmm... she is looking more and more like an ex.
Maybe my subconscious is taking the wheel...
>>8171248They lookalike I even had a gen that confused me at first too.
1.5 tile still the GOAT
Any celebrity you degens want a LoRA on that doesn't have a proper yet or none at all?
>>8176393Olympia Ivleva (formerly of the band Little Big, now an actress)Andre/Andreja PejicJosephine BakerSpider_Monarch aka Cursed_ArachnidNichelle Nichols aka Lieutenant Uhura (60s era)
>>8176393>>8176424Also Tina Louise aka Ginger Grant from Gilligan's Island
>>8176381>>8176424>>8176482This is wonderful.
>>8176497>>8176482>>8176440no wonder she became a nun
Anyone can give her both rectal and vaginal prolapse?
It's somewhat a /pol/ thing, a girl atracting some white masters sexual interest
Happy May Day!
>>8177281
Requesting a real life-ification of pic related.
>>8167007Any rules on requesting things of random obscure Youtubers?
>>8179220you asked for it
How's the realistic slop thread feeling tonight lads
>>8177389i am flattered that you reposted one of my gens and horrified that you took it to its logical conclusion. only my specific amount of implied sexual violence is acceptable, you degenerate
>>8179617nicely done
>>8178661Catbox? Very pretty.
Thoughts? just running img2img with a pit lora
>>8180209A fine bimbo example.
>>8179600This is beautiful. Holy shit, this is good.
>>8170095that's classic style cartoonand it's adult, without questionthough what is your complaint, Sire?
Is there still no checkpoint that produces girls without plastic look?
Small size bc dezgoposting from work...
>>8180619>plastic look?
>>8180884No gen in this thread looks real
>>8180401this is the realistic thread
>>8170095excuse me sir, this is the adult cartoon board. cartoons most definitely belong here, even if this thread focuses towards more real. if it's only supposed to be realistic without a hint of cartoon, this thread should be deleted. glad I could clear things up for you.
>>8180906There was an explosion of development and then everything went to zero out of nowhere.
>>8180960Not all realism is photography.
>>8180619>(render:1.3) in negativesepicrealism is ok with the right prompts (though certain prompts, like film grain or film photo give sameface). There are also some XL models that work pretty well. Albedobase can be pretty good if you boomer prompt on low cfg. Lowering CFG to like 3 is pretty good for this too.
Requesting Liz Gilles please
Bump
>>8182242Moreeeeeeeee
>>8180228>>8180220Get your ass to Mars.
>>8180209Nice, love the style, smoking hot bimbo
>>8182929Very, very, nice work.
Can anything be done with this pic or it doesn't work that way?s0knak
>>8182975You just use a Clarke LoRA.
>>8179600The requester of >>8179220 again. Can I request another race?I would love to see a black woman. I'll also need a white woman to brag.
Still grappling with how to de-same their faces.
>>8183064(different girls, different faces, different facial features) will sometimes work, but there’s always the inpainting slog
>>8182915>>8182918>>8182921>>8182925>>8182929>>8182949holy shit that res - do you heat your house with your GPU?
>>8183157not the guy you're replying to, but my 3080 will definitely heat up the room at least 7 degrees with continuous use
>>8183062>>8183063Very nice.
gonna gen tonight, please post requests which I will probably not do>>8183064(clones, same, similar:1.3) in negatives>>8182975generally, outpainting is what you're looking for, although i hope that's not someone you know. making deepfakes of people you know is absolute degeneracy and is bad for your brain>>8183067>>8183065these are good, can you catbox>>8169961>>8169976lol but also you need jesus
>>8184200Dude that is not someone i know that is Emilia Clarke :)))) i was just wondering if someone can do some nudity with this pic
>>8184439
>>8182929Sweet Jesus what model are you running
If I may make a request; A bunch of people [both male and female] nude working in a very large greenhouse.Alternately , an outdoor cafe where every-one is naked.
>>8182420A Total Recall reference?Nice , I like it.
MAY THE 4TH BE WITH YOU
>>8185771
>>8181474But you have to take a photo of it to share it on the innernet
>>8186337seems easier to just upload the image than to take a photo of it, but okay...
Someone posted a Gillian Anderson/Scully X-Files prompt a month or two back, and I was able to get some decentish results from on Civitai with some tinkering. Less Gillian Anderson, more her stunt double/porn parody actress.But I'm not fully happy with them, and I feel like if they could have one more pass to fix a few of the minor issues they'd be perfect. Lemme show you a few and see what you all think of them.
Like, should I keep going or stop here?
>>8186115Thank you.Do you want more ideas or would that be considered rude?
>>8187161This is a fairly slow thread. It's normal for it to have gaps of 1-4 hours between posts sometimes.Anyway the skin tones look a bit messy with some discoloration. idk if that's from the data set using old screenshots, a VAE issue, or just multiple lora causing it. If you're going for the vintage 90s screenshot look on purpose then that's fine.They don't particularly look like Gillian Anderson to me, tho. What model are you using?
>>8187240Civitai basically told me not to use anyone from real life so I shrugged my shoulders and adapted. I was cycling in LORA's and models left and right just seeing what worked and what didn't, since this was back when they didn't charge you for it.For instance, I think >>8187143was using SDXL unstable diffusers as the model, then had the LORA's Realistic Barbie Girls (set at 0.70) and of all things: Tgirls (XL and SD1.5) set at 0.60. With the prompt:>((huge breasts, cleavage)), 20 year old Dana Scully S7TI, as gorgeous young busty agent scully, sexy FBI special agent,((being turned into a muscle bimbo.)) ((Breast expansion)) (surprised expression),photorealistic, detailed face, 20 years old, ginger hair in a short bob, cinematic, (huge breasts, cleavage), smooth flawless skin, detailed hair, in the cluttered x-files basement office, cluttered dark office, dramatic lighting, hypnotised pose, brainwashing, hypnosis, shadows, ufos, aliens, paranormal phenomena, 1990s style, 90s, depth of field, (NSFW)But obviously >>8187136 and >>8187147 look to have had different prompt and models.
Requesting;Eight ponygirls pulling a marriage carriage done up in fancy ribbons, bows, and streamer cans.Underneath are the words "Always the bridle-mates never the bride"Another idea , a team of ponygirls pulling a stage coach through a western setting.
I'm tempted to try to get into this just to make some pics of Kate Bush in her younger days, she was a super cutie
>>8189223Based idea. Do it!
>>8189056Catbox? Incredible work
Is this thread vanilla shit only? I'm a /d/egenerate and wanted to see how real I could make some other the more uh "niche" stuff. Turns out it can get pretty real. Cant post that at /d/ though since they only accept anime.(Its not futa.)
>>8190175I don't know what catbox is, but I img2img a blender render with RealismEngineSDXL and the all-in-one POV LoRA on civitai>amateur iPhone photo, Caucasian woman wearing flight attendant uniform and sitting on airplane seat, open mouth tongue out, man standing over woman, hand stroking penis, in a large airplane
>>8190375as long as it's not v& shit, yiff, grey background slop, or 512x768 with deformed faces, go for it and we'll yell at you if it's bad.>>8190388is there a reason you use that particular model? I do a lot of img2img too for upscaling and it works but i'm always looking for a model with better faces and skin tone
Okay, I’m a total idiot who wants to get into making this stuff. Where do I start?
>>8190984>>8167007>>Guides
>>8190994NTA but I was about to ask, i see some guides say to install python and git first but these guides on this thread don't, I feel like I should do it anyway? Pic unrelated, I made it in a browser hosted AI Art generator
>>8191375Im sorry, Im a goddamn retard that should learn to read, I see the 2nd method mentions all that. I guess I'll just try the first one first lel.
>>8167302god, this fucking rules
Transsexuality bug, anyone got this already?
I gave this a shot, downloaded stable diffusion and put in some models and I just couldn't get anything that looked good, even if I copied people's prompts. Im gonna completely delete it and try installing it again. It's much easier to use a website like civitai but you only get a limited number of generations so once you got the ball rolling i cant go further.
>>8195010what models are you using? remember that dimensions are a major part of your gen. Also some models use clip skip 2. What happens if you plug in to PNG info?>>8192269thanks. i seem to have lost that file but here's a similar one>https://files.catbox.moe/moh65m.jpeg
>>8195725I downloaded all the models lol. I literally just browsed the AI model website and took what looked good. As I understand it, some need other things to work, and some are not models but embeds or other such things. I thought I understood, but I guess I didn't.For the sake of simplicity and to keep it on topic, Lets say i try CyberRealistic since it is based on SD 1.5 and thats the most basic model (i think) Im going to try using the prompt for this image (pic related) and see what happens. Also, I dont know what clip skip 2 is but i will look into it.
>>8195763lol. okay better than before. I think I know what i did wrong last time (maybe) I was just using the standard Stable Diffusion model. I've seen there are some models(?) that are just close ups of celeb faces, Im guessing those don't work as good for the full body stuff? Also should I just join their discord and bug them with all my questions?
>>8195813Yes bsae SD1.5 is horrible. The later models are much, much improved. Those celeb faces closeups are LoRAs; basically addons or patches for a specific thing you want. The problem is that, in particular with celebs, might done with poor reference, so getting more than an expresion is hard.
>>8196819Combination of celebs can work too for gettingstableidentities.If the training set is small it may be possible that one particular identity get associated with an expression or a clothing style for example, that could make control difficult. You also have a reduction of expressivity possibilities with damboru tagging, wich somewhat want to go from sentence understanding to bag of words.
>>8196819On another forum, I saw some guy say he does this to get his realistic images"Stable Diffusion with RealVisXL 4.0. With InstantID for celebs"What is Instant ID? I didn't see that in the available models, is it some kind of extention?
>>8197838WHY THE FUCK is the current generation so internet illiterate?I literally typed instandID in a search engine and it was THE FIRST FUCKING CHOICEhttps://github.com/InstantID/InstantID
>>8198306It's less about illiteracy and more about incompetence and laziness I think.
>>8198347the funny thing is you just know they wont even know what to do with that link and continue begging
>>8198745
>>8198749>>8198749
>>8198794
>>8198799
>>8198749holy fuck any more of melina?
>>8198802
>>8198809Only if you don't misspell her name, heretic
>>8198816
>>8198810
>>8198820
>>8198824
>>8198828
>>8198306>>8198347>>8198358I was born in the very early 90s, Im not unfamiliar with computers or the internet, I have been on 4chinz since 08. I have been using computers since windows 95 and Mac OS 7 What I am unfamiliar with is the post instagram era internet layout of fucking everything. It especially took me some time to understand how the fuck github even works, when you come from a world where download pages were straightforward and client facing and not fucking layed out like college research papers you have to reach past the abstract to find the instruction on how to download something. I am not a programmer, i do not care that this thing or that thing is a fork of some other fucking thing, I am simply asking how to use something so that I might participate. Anyway here is another pic I made last night thanks to this anon >>8195725 for the help
>>8195763>>8199181ok so:there are models, which is the big 2-6 GB file that makes your image. Then there are LORAs and Embeds, which are basically two different ways of adding "things" that the model doesn't know as the text you type to get it. You always need a model to gen; you don't always need a lora or an embed. If you see something in a prompt separated by <> that's a Lora. Both Loras and models are .safetensors files. There are a few general classifications of models. SD 1.5, SD 2.0, and Pony are the major ones. There's debate and use-cases over which to use. My personal opinion is that my favorite models are 2.0 but controlnet (especially tile, the GOAT coom tool) is better in 1.5. Loras for 2.0 won't work on 1.5 and vice versa. Pony is for coom but the photorealism isn't good.At the end of the day, I recommend you leave this thread immediately and do not make porn with SD because it is extremely addictive.
>>8200050Thank you for the help anon, it is very fun to mess around with and it is addictive because you get the instant gratification and also "just one more generation..." type deal. I can see how it could take time away from self improvement. I'll just use it to make hot red heads.
>>8200232A point I´d like to make on >>8200050 : In practice, the used models (and LoRAs) are 1.5 and SDXL (of which Pony is the killer app).
Would you let her haunt you?
>>8197212Asking the important questionshttps://youtu.be/pkMoIMBBEXQ?feature=shared
>>8201822She's hot.
>>8198816>>8198818THANKS
>>8202185Baked a fresh lora of her, I can make as many as you want
>>8200050Lmao that last sentence. 1000% facts
>>8202044Thanks anon, creepy bitches are the best.I originally wanted to make a haunted ghost girl but failed, nevertheless I'm happy with the result.
>>8183157I gen at a lower res and just upscale - it's a little fuzzy but I'm going for a photo look>>8184464halcyonSDXL with a wall of text dynamic prompt for freckled redheads - pony w/ refiner / img2img can't do the level of realism I'm trying for
and I just realized the irony in saying I'm trying for realism along with a sci-fi themed post
I won't post too many but man I've got so many 1-3girl redheads
these are rare without lora, and imo lora usually fuck everything else up
