Playing with a kindly AI

[lionbadger]

and discovering that AIs have gotten really talented

[lionbadger]

oh, I put a bunch in my deviantart gallery

https://www.deviantart.com/lionbadger/gallery

[lionbadger]

oh,oh, I am, obviously, biased to blondes but did go mildly off piste

[PartsUnknown]

and discovering that AIs have gotten really talented

That gold/white getup looks amazing! Producers, get this costume made, and find an actress to fill it!

[Supergirllegend]

Damn, what the hell am I doing wrong. These are amazing. Any chance of making a full body character of Helen Slater’s Supergirl?

[lionbadger]

I think the Lexica Aperature models are geared to the genre and are doing the heavy lifting

it's just something I'm playing around with so not really into doing requests but, I already ran a supergirl test and got these

b14a1d58-62c6-4cda-ab57-b44063d08b95.png (4.7 MiB) Viewed 17814 times

26161759-9eaf-417a-8fbd-34303981da66.png (4.98 MiB) Viewed 17814 times

supergirl_sd1_by_lionbadger_dfnxdmx-fullview.jpg (58.73 KiB) Viewed 17814 times

[tallyho]

Fantastic job! Not sure how these sneaked under my radar but glad I found them now

[sneakly]

My playing around with AI (I burned through the trial renders) brought up a recurrent problem. The hands often came up weird with odd fingers and what-not.

[ThatOnePervert]

My playing around with AI (I burned through the trial renders) brought up a recurrent problem. The hands often came up weird with odd fingers and what-not.

Hands seem to be a real problem for AI art, even real artist struggle with them and with a smaller sample size of good hands to draw from it will probably be a while before some of these programs make them work.

It reminds me of some old lore about succubi. The idea was that a succubus would always have one inhuman trait, like a duck's foot or a thin tail because God would always give you a way to notice when a demon was trying to seduce you. Likewise you can always tell AI art from real art by looking for the hands.

[lionbadger]

My playing around with AI (I burned through the trial renders) brought up a recurrent problem. The hands often came up weird with odd fingers and what-not.

yeah, for every good one I get there are 6 that look like gremlins dipped in water

[sugarcoater]

So stupid question here, but is this AI tech something available to any of us here? And if so, how do we acquire the program? Just thinking of trying my hand at art to supplement my story.

[Mr. X]

So stupid question here, but is this AI tech something available to any of us here? And if so, how do we acquire the program? Just thinking of trying my hand at art to supplement my story.

Yes available to anyone.


Here's a video that helps you install it. This guy's channel discusses how to use it.

[DonShip]

These AI pictures are rather inspiring. I'm creating a new OC super heroine, and I want to do something other than my usual bustier, thong, and thigh boot costume. But I haven't decided on a heroine name, and that will inspire the costume style and color.

But still, very inspiring pictures.

[Mr. X]

These AI pictures are rather inspiring. I'm creating a new OC super heroine, and I want to do something other than my usual bustier, thong, and thigh boot costume. But I haven't decided on a heroine name, and that will inspire the costume style and color.

But still, very inspiring pictures.

The description is the key. The AI knows Wonder Woman. It does not know Ms. Americana. So when you say "Wonder Woman" anywhere that becomes the theme.

[argento]

Very pretty. They all have a diamond face and almond-shaped eyes.

[argento]

and discovering that AIs have gotten really talented





The first one is the one I like the most

[brdiy]

Being as we're discussing AI, how do you feel about superheroine stories written by (or with the assistance of) AI?

I have been playing around and while I can't really come up with something too explicit, I would say the results have been satisfying (and mildly addictive ) so far.

[HeroineTard]

I have used "AI Dungeon" to create some perverted Superheroine related text adventures.

I'll say that the paid subscription with the advanced AI comes up with some really interesting stuff. I can totally understand how addictive it can be.

[lionbadger]

(can't beleive I put this in the wrong thread!)
anyway, as I was saying, this has turned into my Sunday night hobby

tethys 19.png (527.24 KiB) Viewed 17206 times

tethys 12.png (564.88 KiB) Viewed 17206 times

metallica 3.png (606.25 KiB) Viewed 17206 times

valk 19.png (513.76 KiB) Viewed 17206 times

[HeroineTard]

(can't beleive I put this in the wrong thread!)
anyway, as I was saying, this has turned into my Sunday night hobby

tethys 19.png

tethys 12.png

metallica 3.png

valk 19.png

Nice! What do you use to generate these AI images?

[lionbadger]

Stable Diffusion - Dano of Superheroinecentral fame posted some stuff in December and I was sort of inspired - he's posted load of stuff since on his deviantart and at shc

playing with different models to generate, Lexica has a good free for so many uses one you could give a go

[lionbadger]

oh and because I know some of you like that underwater stuff

valk 40.png (563.83 KiB) Viewed 17174 times

[Mr. X]

Next step - Text to AI generated video

[Mr. X]

Some fun using Melissa for Supergirl

1155.jpg (444.21 KiB) Viewed 17142 times

[Dogfish]

There's a rumour around that the latest version of one of the main art AIs, I think it's Stable Diffusion, can do hands and fingers. It took a while.

It's interesting to me how this was a blindspot, and it comes down to how they draw. They're not drawing, basically the AI doesn't know how many fingers a human has. That's not how it works. It's weird.

[Mr. X]

There's a rumour around that the latest version of one of the main art AIs, I think it's Stable Diffusion, can do hands and fingers. It took a while.

It's interesting to me how this was a blindspot, and it comes down to how they draw. They're not drawing, basically the AI doesn't know how many fingers a human has. That's not how it works. It's weird.

Yes SD starts with a noise pattern pic similar to what you see on an old TV with no signal. It then starts pattern matching based on criteria slowly evolving that white noise into a pic. SD now has the ability to use stick figures to pose a figure and a redraw pencil feature to redraw areas.

00008-217385672.png (853.56 KiB) Viewed 17088 times

I did this one with a stick figure with hands over the head.

[lionbadger]

SD now has the ability to use stick figures to pose a figure and a redraw pencil feature to redraw areas.

I did this one with a stick figure with hands over the head.

oooh, interesting, are you doing that via img 2 img?

[Mr. X]

Adobe's entry into the market place. A lot easier to use.

[lionbadger]

some sunday shenanigans (basically me trying to convince the AI to make a 'superhero eye mask', with, err mixed results)

valk 65b.png (548.53 KiB) Viewed 16874 times

valk 65.png (546.99 KiB) Viewed 16874 times

valk 62.png (544.19 KiB) Viewed 16874 times

valk 56.png (526.01 KiB) Viewed 16874 times

[Mr. X]

Some AI art posted at the Henchmen's Lounge.
https://subscribestar.adult/dangerbabecentral

[Jokey1234]

I lurk around here often but I noticed all of your AI art and got inspired to try. I've only really started earlier this week so my AI art hasn't looked great.

I apologize mine isn't as classy either

Hopefully I posted this correctly, I haven't posted images on this forum before either.

[lionbadger]

playing around with slightly more abstract concepts

la tigra2.png (560.23 KiB) Viewed 16591 times

la tigra1.png (563.6 KiB) Viewed 16591 times

[lionbadger]

same idea, different models

la tigra5.png (516.68 KiB) Viewed 16585 times

la tigra6.png (545.23 KiB) Viewed 16585 times

[sneakly]

I am curious if you can get character continuity from these programs, where you can fix a character, setting and then have the computer render different scenes within the setting. For example it creates a batgirl and riddler in his hideout and storyboard the entire sequence with descriptions of the actions while using the same batgirl and Riddler over and over.

[Mr. X]

I am curious if you can get character continuity from these programs, where you can fix a character, setting and then have the computer render different scenes within the setting. For example it creates a batgirl and riddler in his hideout and storyboard the entire sequence with descriptions of the actions while using the same batgirl and Riddler over and over.

Not right now but its getting fairly close. You can to an extent if you use say Poser or Blender to create depth map templates then image 2 image to get it to AI-ify the renders. But that consistency should be coming. There is a new chatgpt that allows you to create a character template with characteristics that you can feed into the AI for story writing. That probably will merge at some point. Again this is all new but it changes every month. I'd say in less than 3 months you'll see consistent pics of the same character.

[Supergirllegend]

We have the technology gentleman

[lionbadger]

couple more

spygah3.png (575.68 KiB) Viewed 16506 times

spygah1.png (614.78 KiB) Viewed 16506 times

[Mr. X]

One issue I see now is a lot of products sold on places like DLSite are CG collections in which someone fired off 20-30 pics and sold them. Basically there isn't much skill in that given they don't even touch them up like removing extra fingers etc. I do post some but I'm trying to do original work and the AI are just fluff. At some point I think people will catch on that, while these pics look good, they are just pics right now, no story.

[ThatOnePervert]

One issue I see now is a lot of products sold on places like DLSite are CG collections in which someone fired off 20-30 pics and sold them. Basically there isn't much skill in that given they don't even touch them up like removing extra fingers etc. I do post some but I'm trying to do original work and the AI are just fluff. At some point I think people will catch on that, while these pics look good, they are just pics right now, no story.

That's a feeling I have with a lot of AI art. Like there's already moral issues that I'm not gonna argue because I'm not that familiar with the process, but on the whole most AI generated heroine pics are just some woman in a vague costume staring forward without any substance or peril, the few people I've seen get more daring usually cross the line into nightmare fuel, where you'll see some tentacle play, but the tentacles are growing out of the heroine or they tried some bondage and now the heroine's legs just meld together instead of ending in feet. Some of the base concepts are hot, but the AI still leaves a lot to be desired and the fact that people are unironically trying to charge for these things feels half horrifying and half laughable.

[Mr. X]

One issue I see now is a lot of products sold on places like DLSite are CG collections in which someone fired off 20-30 pics and sold them. Basically there isn't much skill in that given they don't even touch them up like removing extra fingers etc. I do post some but I'm trying to do original work and the AI are just fluff. At some point I think people will catch on that, while these pics look good, they are just pics right now, no story.

That's a feeling I have with a lot of AI art. Like there's already moral issues that I'm not gonna argue because I'm not that familiar with the process, but on the whole most AI generated heroine pics are just some woman in a vague costume staring forward without any substance or peril, the few people I've seen get more daring usually cross the line into nightmare fuel, where you'll see some tentacle play, but the tentacles are growing out of the heroine or they tried some bondage and now the heroine's legs just meld together instead of ending in feet. Some of the base concepts are hot, but the AI still leaves a lot to be desired and the fact that people are unironically trying to charge for these things feels half horrifying and half laughable.

Yes right now, to get some pose etc its a herculean effort. DanO at SHC pulls off some great stuff but I know he has to generate displacement maps then use a process of img2img then go back in and repaint areas. So its not just "snap" you get art. The stuff I make is like you say, some standing pose with no peril or in some stock poses. In fact all you get is some costume ideas. It will get to be tech that creates what you want however I don't think people should worry about putting artists out of work cause describing what you want is a trick in itself. I find a lot of people don't exactly know how to describe what turns them on, they just know it when they see it. You can hand someone a pencil and paper and they have all the tools to make a comic book but they don't.

[ThatOnePervert]

Yes right now, to get some pose etc its a herculean effort. DanO at SHC pulls off some great stuff but I know he has to generate displacement maps then use a process of img2img then go back in and repaint areas. So its not just "snap" you get art. The stuff I make is like you say, some standing pose with no peril or in some stock poses. In fact all you get is some costume ideas. It will get to be tech that creates what you want however I don't think people should worry about putting artists out of work cause describing what you want is a trick in itself. I find a lot of people don't exactly know how to describe what turns them on, they just know it when they see it. You can hand someone a pencil and paper and they have all the tools to make a comic book but they don't.

It's not that I think all artist would be replaced by AI, but I think a lot of people work on the idea of "good enough." AI doesn't have to be perfect or recreate their fetishes to a T, but as long as it's cheaper than hiring an actual artist and it's good enough for whatever a person needs or wants in the moment then they'll take it. And I think enough people think that way to lower the amount of money that can be made from art and practice opportunities and artist gets through commissions.

[Mr. X]

This is what DanO is achieving with depth map renders and image to image.

[batgirl1969]

But why are the breasts always so......big? Is there no love for smaller petite built heroines

[sneakly]

But why are the breasts always so......big? Is there no love for smaller petite built heroines

You realize it learned human anatomy from pictures on the Internet. Probably lucky it doesn’t think a strap-on dildo is an inherent part of the female anatomy.

[Mr. X]

But why are the breasts always so......big? Is there no love for smaller petite built heroines

You realize it learned human anatomy from pictures on the Internet. Probably lucky it doesn’t think a strap-on dildo is an inherent part of the female anatomy.

His pic was done with an original pic as the template. So yes his original pic has giant tits.

[Mr. X]

But why are the breasts always so......big? Is there no love for smaller petite built heroines

You realize it learned human anatomy from pictures on the Internet. Probably lucky it doesn’t think a strap-on dildo is an inherent part of the female anatomy.

His pic was done with an original pic as the template. So yes his original pic has giant tits.

Here is the original he used to tell the AI how to make the pic.

[DonShip]

This is what DanO is achieving with depth map renders and image to image.

An amazing image of Power Girl, but her face is above and beyond. Wow.

[lionbadger]

Here is the original he used to tell the AI how to make the pic.

That's quite clever, make an img in poser and ask the AI to re do it in a certain style

[lionbadger]

did a quick test on the img2img method, I would call it a "qualified success" and I need to refine a bit (or be less ambitious) but the principle seems to work after all of 20 mins messing around

valkyrie i2i1.png (560.1 KiB) Viewed 16139 times

[lionbadger]

we getting there now!

valkyrie i2i2.png (535.56 KiB) Viewed 16131 times