AI generated art experiences for NSFW content

bobjones9792

Member
May 27, 2017
263
732
So many of you have seen the talk about AI generated art. There's been a lot of discussion, even on major new networks, about what AI art means for the future, the ethics involved, and the potential legal issues down the road.

I'm not going to be talking about any of that.

Instead, I'm going to give those of you who are curious a heads-up on the state of NSFW porn generation, specifically actual sex and not just nudes. For background, I'm using Stable Diffusion Automatic1111 webui version, which I believe is built on the older SD1.5 standard. I've been playing around with it for about a full day now, setting it up (which was annoying and not for the technically challenged), reading up on creating good prompts, experimenting with different models (which are basically the "pool" the AI draws its information and image data from), and creating images while experimenting with the prompts. I tried the Hassan, URPM, Smirking Face, Grapefruit, and Waifu models. I've been using extensive negative prompts, and relatively simple positive ones. So how was it?

Its not that good. Don't delete your porn collection just yet.

Its pretty obvious that the AI can't generate stuff it has little exposure to. When I tried a BJ picture under a desk, it made a few images with a desk near them, when it bothered at all. It is very good for the most part at swapping in small details, like a choker, on people, but it struggles greatly with any kind of fine details. It also has a limb problem, especially a man's third limb, as it tends to do things like making it grow double in size, float in mid-air away from the body, and duplicate it in different spots. It also struggles with actions, often ignoring explicit prompts like "kneeling". These problems applied to both realistic and hentai style art, however the hentai style was far worse. It seems that hentai models are even worse about correctly displaying dicks then RL ones, often showing a heavily censored dick that doesn't even look like a dick floating in front of the girls while they poorly do, well, something with it.

The biggest issue though, is it just flat out ignores parts of the prompt. I typed "girl getting fucked doggy style", and not one of the images showed a doggy scene. Prompt creation is a mess because you sometimes need to know the exact words to trigger creation.

That said, it does do a good job at creating good looking girls standing around or posing. Swapping in and out details like hair color and simple accessories is easy and mostly goes fine. Its great at backgrounds, and portraits. Its basically decent at anything that involves not much happening, provided you don't expect it to read your mind.

Give it a year, and it might actually be a serious threat to traditional image porn. Right now though? Not really.
 

Doorknob22

Super Moderator
Moderator
Game Developer
Nov 3, 2017
2,391
5,809
Give it a year, and it might actually be a serious threat to traditional image porn. Right now though? Not really.
In theory, AI and porn could have mind blowing possibilities: no longer praying and hoping that your favorite smut author/artist will issue something which tickles your darkest fancies and no more worrying lest someone finds your shady browsing history. The AI will cater to your darkest desires with no moral boundaries, just hop on its back and ride the vastness of Smutland to wherever you heart desires.

Reality, however, is different. Trying to get an AI storyteller to tell a decent pervy story is akin to getting an extremely absent minded person through an obstacle course: eventually you might get to the end but the amount of retries will leave you exhausted. On the visual front, the AI will produce horrific nightmares and passable smut with similar enthusiasm with little ability to differentiate between the two, not to mention its problem with producing consistent images of the same characters if you wish to use it for comics or games.

Will the AI improve? Maybe. Humans have a tendency to expect technology to improve in a linear manner but history shows that improvements in technology are erratic. The AI we now have might be "as good as it gets" for a long time.
 
Last edited:

anne O'nymous

I'm not grumpy, I'm just coded that way.
Modder
Donor
Respected User
Jun 10, 2017
10,957
16,192
There's been a lot of discussion, even on major new networks, [...]
Including here. And I have to say that it starts to be annoying to see a new AI thread pop up every week or so.


Will the AI improve? Maybe. Humans have a tendency to expect technology to improve in a linear manner but history shows that improvements in technology are erratic. The AI we now have might be "as good as it gets" for a long time.
This apply even more to AI since the main problem lie inside their black box, and by definition no one know, nor really understand, how they works.
Improving AI was relatively easy until now, because the goal was just to make them works, yet it needed near to 60 years to reach this result. But now that what is needed is fine tuning, it's something radically different.
There's no way to know why they suddenly decided to put 7 fingers to the hand they drawn. Their switches made them reach to this conclusion, but why and how, no one know and, as long as we will rely on black boxes, no one can know. What mean that no one can solve this issue.
The next AI will have even more switches, what is expected to make them even more precise. But they'll still not be more reliable, just weird in a different way.
 

Guntag

Active Member
Donor
May 3, 2017
702
857
I know nothing about all this, but out of curiosity, I watched this video a few hours ago. I doubt it could be used for NSFW games since it's mostly about faces and seems too complicated, but it was an interesting watch.

 
  • Like
Reactions: Brine_Basilisk

zenshiny

Newbie
Dec 31, 2016
56
92
The AI we now have might be "as good as it gets" for a long time.
I think we still have some ways to go before we reach that point. ChatGPT's got the hype right now, but Google's got an even bigger model, which they'll use to make their own version of ChatGPT combined with Google Translate. MTL poetry might go from avant-garde to utterly pedestrian!

The limits of current training methods--as far as I understand, anyway--is having massive amounts of data to train the model with, and having the hardware to train models on that data in a reasonable timeframe. I imagine that the tech giants all have immense hoards of data they've barely tapped into.

I used to see buzzwordy articles on sites like Wired that said, "Data is the new oil!" and I'd just roll my eyes, because those articles would apply the phrase to advertising and marketing and getting more customers for businesses. Now I'm imagining tech giant baronies creating an OPEC of data to fuel AI engines.
 

bobjones9792

Member
May 27, 2017
263
732
Including here. And I have to say that it starts to be annoying to see a new AI thread pop up every week or so.
The technology has taken off recently and is rapidly changing at this point. There's talk about upcoming performance improvements that will allow for real time generation, or at least allow for single images done in trivial amounts of time. New models are coming out constantly, many of them focused on specific content. Its understandable people want to talk about new stuff.

That said, a lot of discussion seems focused more on possibilities, which is why I made my post about the hands on experience. I was really excited at first, and its still a neat toy, but ultimately AI art generation is too much work for the return right now, at least for anything but one offs with low standards.
 

lancelotdulak

Active Member
Nov 7, 2018
556
557
So.. anyway... started programming 40 years ago... because i wanted to do ai... anyway... i have stable diffusion with automatic1111 installed. About 6 checkpoints (thing MASSIVE databases of pictures). Ive generated some pretty sweet erotic stuff.. ill post a few. But saying "it isnt replacing porn anytime soon etc" when youve played with it 2 hours is like saying you cant create photorealistic content with daz after an hour.. or a week. Oh also there are companies actually starting to do it.

Guided creation is harder and requires more work and expertise. You use tools that people just tossing in prompts dont even know exist. And the KEY is the checkpoint. The big checkpoints are built from pictures that specifically exclude nsfw. We'll need people doing a huge amount of work building checkpoints for nsfw .. which is already happening. BTW you folks thinking about investing money into these big corporate ai's... dont. Theyre all hyper obsessed about excluding nsfw and any content they dont like (a lot of which we wouldnt like). Porn built the internet folks. Those companies are going to go byebye (especially as their creations will be based on others work and.. actionable). Each of these took < 20 seconds with a 3080ti..
I dont like the generated sex photos yet.. i need to do more work and need better checkpoints etc
 

Hadley

Well-Known Member
Sep 18, 2017
1,047
1,892
I tried some and its really hit and miss. The problem you can't really re-create the same person over and over again in different animations unless you already have a base you can use.

When we get some really good models that are trained for nudity/porn it would obviously help a lot.

But so far the tech isn't there yet unless you just need some random decent images.
 

Writefuck

Newbie
Game Developer
Dec 7, 2018
22
61
Give it a year, and it might actually be a serious threat to traditional image porn. Right now though? Not really.
That's the thing about AI. It gets better at doing things, but it doesn't even get (permanently) worse at doing things. As it learns and we get better at teaching it how to learn, its ability to do things will only improve and improve. Right now I feel that AI is generally kind of bad at making art, but that hardly matters since it's all uphill from here. It's going to get better.

(insert Kent Brockman quote about welcoming our new yadda-yadda overlords)
 
  • Like
Reactions: lancelotdulak

MissFortune

I Was Once, Possibly, Maybe, Perhaps… A Harem King
Respected User
Game Developer
Aug 17, 2019
5,375
8,643
Right now I feel that AI is generally kind of bad at making art, but that hardly matters since it's all uphill from here. It's going to get better.
There's already artists and companies suing AI Art generators. Depending on how that goes, it could massively stomp out any future AI art has for a while, to a point where it's for fun and illegal for any commercial use. Stealing images for training without any attribution, payment, credit, or even the licenses to do so is a huge copyright violation, and the copyright registration being pulled/declined on that comic book/graphic novel is another kick in the face toward AI art.

Until that battle is out of the way, there's no real future for AI art in any meaningful form. If and when games with AI start hitting up Steam, they'll be quick to ban it in fear of potential lawsuits from either artists or the likenesses of models created by the generators. Patreon will likely follow suit. Even if SS sticks around for it, that more or less will stop anyone from doing it outside of passion projects.

Edit: Gotta love it when people facepalm you for telling the truth lol.
 
Last edited:

Trickstar

Well-Known Member
Jun 2, 2018
1,476
5,654
I have actually already seen several amazing arts for hentai already that is ai. It's honestly incredible what people can already do with it. With more advancements and more practice with the tools I honestly think ai will definitely be very strong in the future for hentai and more.
 

Writefuck

Newbie
Game Developer
Dec 7, 2018
22
61
Stealing images for training without any attribution, payment, credit, or even the licenses to do so is a huge copyright violation
I'm not totally up to date on any legal stuff, but has a court decided this definitively yet? Whether or not the act is lawful means almost nothing until a judge or a police officer enforces that law.
 

MissFortune

I Was Once, Possibly, Maybe, Perhaps… A Harem King
Respected User
Game Developer
Aug 17, 2019
5,375
8,643
I'm not totally up to date on any legal stuff, but has a court decided this definitively yet? Whether or not the act is lawful means almost nothing until a judge or a police officer enforces that law.
Well, the has already lost copyright. As I said in the quote post, the landscape will likely be heavily impacted by , which only was brought to the headlines a few weeks ago (at most.). The datasets themselves are stolen/scraped from websites and used without any level of permission and often commercially. I think that sort of speaks for itself.

While there's no copyright enforcement for infringement currently, it means a lot from an artistic perspective. Would someone pay/commission for AI art that anyone else can take for free and use without punishment? No. Would anyone make a game with AI art knowing the said work could be stolen and used elsewhere? Probably not. Until there's a profitable copyright system that works for all sides, AI is nothing more than a toy.
 
  • Like
Reactions: lancelotdulak

♍VoidTraveler

Forum Fanatic
Apr 14, 2021
5,786
15,405
So.. anyway... started programming 40 years ago... because i wanted to do ai... anyway... i have stable diffusion with automatic1111 installed. About 6 checkpoints (thing MASSIVE databases of pictures). Ive generated some pretty sweet erotic stuff.. ill post a few. But saying "it isnt replacing porn anytime soon etc" when youve played with it 2 hours is like saying you cant create photorealistic content with daz after an hour.. or a week. Oh also there are companies actually starting to do it.

Guided creation is harder and requires more work and expertise. You use tools that people just tossing in prompts dont even know exist. And the KEY is the checkpoint. The big checkpoints are built from pictures that specifically exclude nsfw. We'll need people doing a huge amount of work building checkpoints for nsfw .. which is already happening. BTW you folks thinking about investing money into these big corporate ai's... dont. Theyre all hyper obsessed about excluding nsfw and any content they dont like (a lot of which we wouldnt like). Porn built the internet folks. Those companies are going to go byebye (especially as their creations will be based on others work and.. actionable). Each of these took < 20 seconds with a 3080ti..
I dont like the generated sex photos yet.. i need to do more work and need better checkpoints etc
Could you perhaps consider creating a guide on F95 that would aim to teach devs here how to AI?
I think that would be extremely useful in the long run. Thanks. :whistle::coffee:
 
  • Like
Reactions: Sonydo

lancelotdulak

Active Member
Nov 7, 2018
556
557
Could you perhaps consider creating a guide on F95 that would aim to teach devs here how to AI?
I think that would be extremely useful in the long run. Thanks. :whistle::coffee:
Honestly.. you just need to play with it.. like daz. The learning curve isnt steep. Some tips though:
Install auto1111 or nmkd. Auto1111 is superior. NMKD is easier to install. You can theoretically use a 4 or 5 gig card. My 1070 handles it fine. My 3080ti is crazy fast. You dont need anything near what you need with daz (you need vram..)


Sampling steps around 20. That gives you a good image. Like 3d more is better but with diminishing returns.
Start out with cfg around 7. Higher tends to make it more cartoony depending
ddim is a good all around sampling method
get the gravyblend f22 and protogen checkpoints. More is better. Theres also a porn checkpoint but it's.. blah
Prompts are the secret and it doesnt always work the way you think.
Prompthero.com has some good prompts
Negative prompts matter
Stick with sd 1.4 or 1.5. 2.0 is LESS good
Some good prompt words are photoshoot, 50mm lens, ultrarealistic, highly detailed etc. You have to experiment
Generally id suggest batchcount = 4 (# of pictures with that base model)

And play.. just like you do in daz. The thing that excites me about stable diffusion/ai art is the thing that depresses me. I can play with some keywords and create an incredible model/scene in.. 20 seconds or .. 2 hours. With far less control. Something that would take me days or more in DAZ.. though with daz i have absolute control.

Prompts and the model are the big secret. After that you learn more .. control i guess?
Oh and make sure you have a version of SD with the nsfw filters removed. The devs are all terrified people are going to make things they dont like. People are absolutely going to do that. The cat's out of the bad. AI art may be commercially unviable for a long time (til companies buy up databases) but for us.. we're golden
 

lancelotdulak

Active Member
Nov 7, 2018
556
557
Well, the has already lost copyright. As I said in the quote post, the landscape will likely be heavily impacted by , which only was brought to the headlines a few weeks ago (at most.). The datasets themselves are stolen/scraped from websites and used without any level of permission and often commercially. I think that sort of speaks for itself.

While there's no copyright enforcement for infringement currently, it means a lot from an artistic perspective. Would someone pay/commission for AI art that anyone else can take for free and use without punishment? No. Would anyone make a game with AI art knowing the said work could be stolen and used elsewhere? Probably not. Until there's a profitable copyright system that works for all sides, AI is nothing more than a toy.
I completely forgot about this. Im sure everyone remembers when peta sued a photographer on behalf of a monkey who took a picture that became famous (holy fuck post 2000 just gets weirder) (also fuck peta). Anyway the judges ruling.. which is precedent.. was exactly that. The monkey couldnt hold a copyright... and neither could the photographer because he didnt create it.
There will be workarounds. You can copyright anything you make certain substantial changes to. So long as a human does significant creative post processing work theyll be able to copyright it.
 
  • Like
Reactions: MissFortune

MissFortune

I Was Once, Possibly, Maybe, Perhaps… A Harem King
Respected User
Game Developer
Aug 17, 2019
5,375
8,643
You can copyright anything you make certain substantial changes to. So long as a human does significant creative post processing work theyll be able to copyright it.
But even then, only the changes are protected (e.g. ). While the above is true, it creates a few more layers of complexity to an already complex situation to navigate through. How much of the image needs to be changed before it's considered 'substantial'? How does infringement work with an images that's partially public domain (currently) and partially copyrighted? Does the infringer only pay half of the damages? Do half of those damages go to the artist(s) the AI-assisted artwork used? I think it also needs to be asked - if one's capable of significantly altering an AI image to a degree that it's substantially changed, then why aren't they just making their own? Because it's easier? Faster? Those aren't reasons, just excuses.

The comic linked in the quoted post was denied copyright even with significant changes. Partially because it was a walking lawsuit as it was carrying the clear likeness of a celebrity and thus very vulnerable to damages by either herself or her team, but also because it was machine made/assisted, and thus following current (and obviously archaic, even if I'm personally against AI art.) copyright laws as it wasn't made entirely by a human.
 
  • Like
Reactions: lancelotdulak

M$hot

Member
May 28, 2017
238
371
I just remember seeing Code Bullet vids on YT where he teaches an AI with machine learning on what is and isn't good based on a reward/punishment system, with the goal being the highest point total. So if a move off the tracks gives -5 and making a sharp right to drift the corner properly gives +1, most iterations will try the drift, until you've evolved them beyond that point. And so on and so forth. I have to imagine people will pair this with AI art generators, to essentially grade the results it's giving? Like if you ask for a blonde and you get a redhead, you'd rate the fulfillment of the keyword as 0. If you can teach the AI art generator what you mean with each word you use, it should eventually start yielding better and more stable results right?

Idk, excited to see what the future holds, but since I can barely hold my breath for 2 minutes and this could take until I'm grey, won't be holding my breath.
 

lancelotdulak

Active Member
Nov 7, 2018
556
557
I just remember seeing Code Bullet vids on YT where he teaches an AI with machine learning on what is and isn't good based on a reward/punishment system, with the goal being the highest point total. So if a move off the tracks gives -5 and making a sharp right to drift the corner properly gives +1, most iterations will try the drift, until you've evolved them beyond that point. And so on and so forth. I have to imagine people will pair this with AI art generators, to essentially grade the results it's giving? Like if you ask for a blonde and you get a redhead, you'd rate the fulfillment of the keyword as 0. If you can teach the AI art generator what you mean with each word you use, it should eventually start yielding better and more stable results right?

Idk, excited to see what the future holds, but since I can barely hold my breath for 2 minutes and this could take until I'm grey, won't be holding my breath.
This is how ai is trained and there are levels of this you can do with stable diffusion.. you can train/retrain the model or styles etc etc (different levels of training basically). But training is extremely time extensive.. they paid some kenyan workers like $1.25 an hour to train a 10k image set to teach it to avoid nsfw (which works Badly). The big models use millions of images and to train a GOOD model you need an image with keywords for it and negative keywords for it.. for example girl, standing, beach, bikini, blond, tall etc etc etc along with negatives. It can be done and you can make a custom one yourself. But AI learns like you do.. or a bug or mouse or dog does. It isnt precise. Which is why the right prompts and trial and error are the key. Though i expect 5 years from now theyll be unbelievably good. Oh and nvidia has their own ai.. which hopefully they release but .. i have a feeling will be strictly nsfw etc i just cant see them having the cajones to release it for local use
 

lancelotdulak

Active Member
Nov 7, 2018
556
557
But even then, only the changes are protected (e.g. ). While the above is true, it creates a few more layers of complexity to an already complex situation to navigate through. How much of the image needs to be changed before it's considered 'substantial'? How does infringement work with an images that's partially public domain (currently) and partially copyrighted? Does the infringer only pay half of the damages? Do half of those damages go to the artist(s) the AI-assisted artwork used? I think it also needs to be asked - if one's capable of significantly altering an AI image to a degree that it's substantially changed, then why aren't they just making their own? Because it's easier? Faster? Those aren't reasons, just excuses.

The comic linked in the quoted post was denied copyright even with significant changes. Partially because it was a walking lawsuit as it was carrying the clear likeness of a celebrity and thus very vulnerable to damages by either herself or her team, but also because it was machine made/assisted, and thus following current (and obviously archaic, even if I'm personally against AI art.) copyright laws as it wasn't made entirely by a human.
Theres a copyright lawyer on youtube. .. Leonard French, who makes copyright law actually entertaining and teachs lol. And ya if you watch his channel enough you learn the basics of (C) law and the way the system works right now it will cost you 4 figures to stop an obviously frivolous suit and 5 to 7 figures to fight one that could go either way. And there are no precedents around ai art yet that will take years or decades. If you use it commercially youre going to have to be extremely careful.. which is good news for us daz/poser folks for now i guess lol