3D R Comics

Completely unrelated experimentation

Added 2023-01-13 20:53:48 +0000 UTC

So unrelated to anything.

I was having some fun with AI while waiting on renders because I have been a little curious.. If AI programs can generate images out of nothing. why can't it enhance 3d rendered images?

So I did some research. The way AI images generators work is it pretty much takes an image that just noise and de noises it until it can see what you told it should be seeing, generates some neat results.

I think one of the more interesting aspects of it is you can give it an image and it will tell you what it sees.

I gave it one of the bowling alley scenes and it saw "a woman holding a ball in a room with a brick wall and tables and chairs in the background."

pretty damn accurate.

So I downloaded and learned how to use stable diffuse, after some trial and error I was able to provide it an image to start with, without destroying it, Im still trying to figure out how the control the "noise" image it uses as a base, the process I am using now just kind of keep injecting the image i provided it as its de noising and it uses it as a guide.

this has a habit of destroying the face though, but their is a feature that "restores faces". Now restore I don't think is the correct term to use for it as i think its more deepfake ontop of ai image generation. but I tried to do an reverse lookup on a few of the faces its giving me and cant find them. so idk.

but this is the result of this experiment atm. not consistent, not usable in anyway, but neat.