Viridian's AI Experiments

[Acidtester]

Awesome!
I think her complexion is slightly better than the last one. It contrasts with her hair better, instead everything being kind of the same orangy hue.
This is very hot tho, and she’s looking better and better every day!!

[Viridian]

Jesus Christ, words fail on this one... That shirt..that expression...those...ahem!...
Acid, this is your girl, 1to 10, what do you say...11? 12?
I'm officially beyond any known superlatives now.

Making sure that Susan doesn't miss out.

[Jinn]

Making sure that Susan doesn't miss out.

This one is a winner with me. Love how the pantyhose turned out. Fantastic!

[sixgunzloaded]

Making sure that Susan doesn't miss out.

She and I thank you very much!! Probably me more than she, but she never complains out loud.

Here's something new(?) to try - Can it handle any sort of action shot? For instance, a girl trying to pull herself out with a vine? Maybe looking up in fear, as if to see if the vine is holding?
How about motion blurs? How does it do with those?

Making sure that Susan doesn't miss out.

This one is a winner with me. Love how the pantyhose turned out. Fantastic!

OMG, so do I! Muddy pantyhose and stockings... ~drools~ Even better if they're torn from, say, a long struggle!

[Viridian]

Remake of Acidtester's Officers Down

[Viridian]

Here's something new(?) to try - Can it handle any sort of action shot? For instance, a girl trying to pull herself out with a vine? Maybe looking up in fear, as if to see if the vine is holding?
How about motion blurs? How does it do with those?

I suppose AI could do motion blurs. I've got my prompts set up for still photography, so it's not specifically suitable to capture scenes in motion, but it can generate dynamic images.

5ded1678e0da4dd8b6ab2eba57beee5a.png

As for vines, this is a huge struggle for AI and a big advantage for 3D models. AI doesn't exactly know what to do with hands, so it struggles to connect that the pixels that are meant to be "fingers" are supposedly holding a "vine" (or any prop) in a specific way.

ae10a256c2c04bc2966d5a396100f684.png

I can get around this by using "camera tricks" to hide the action, but you can see here that the hand isn't holding the vine. It just can't.

ed7ecf62bea44fefac172facf0d3c513.png

It's the same thing with ropes, torches, guns, and so on. It's also why I can't generate bondage pictures, such as being tied up and handcuff. The AI just doesn't register how it's supposed to look even with lots of guidance. My "Undermined" Sanders pick is largely luck, basically excluding that part of the picture from the AI render. It's possible, to generate, but requires a lot of specific manual inpainting to get it to work.

Hence the strength of AI is in created posed shots. And one strength of that it is VERY good at pantyhose.

15b442933c8b48ddb14bcd2325cf74a4.png

6fcecb87ad214b45b2cd78a2812cb79f.png

[Boggy Man]

In the last picture, looks like part of her right leg is missing.

[MDLambert]

Love the jungle girl holding the vine. Very cool.

[Viridian]

In the last picture, looks like part of her right leg is missing.

It is, I didn't have time to correct it. Just wanted to show off the pantyhose.

[Viridian]

Let's up the game a little. First, a few basic animations of 2D images using LeiaPix. The last one has some basic AI voices and soundscapes added.