I remember when we were all excited about morphing, fractal generation, and stereograms in the 90s. New visual fx are always exciting for a while before turning commodity. I do wonder if the ML image generation is basically a lineal descendant of that stuff with extra steps
Conversation
Has anyone tried to generate novel fractals or morphs?
“Fractal pentagon”
“Halfway morph of Queen Elizabeth and Oprah” (there was a controversial Benetton ad back in the day of a black QE)
“Stereogram of a horse”
4
6
I’d be surprised if any ML image generator can do autostereoegrams since even basic perspective causes trouble (see my pfp for eg). It’s the pixels of a tesselation perturbed to produce a stereoscopic optical illusion when you look at it in a trick way.
1
1
I stand corrected 😂
Quote Tweet
you can make autostereograms with DALL-E
(relax your eyes and the image will "pop" a bit in 3D)
Show this thread
Replying to
AI going 3d! Or at least stereo. Full 3d will take 3d training data in cad/game databases or metaverse I think
Quote Tweet
yeah, so for everyone saying AI isn't ready or AI art isn't art
i just made a working 3D anaglyph of a t-rex in DALL-E
#dalle
1
8
Replying to
It's more of a pretending to make one than making one - it doesn't produce any coherent image when viewed.


