VISUALS
Custom trained data set based on the work of #geoffreylillemon, #1980shair, #bodybuilders, and a trick of using alpha channels in the data set to create graphic striped lines in the animations.  
Everything trained in #dreambooth then the exported .ckpt file was brought into #deforum.

SOUND
After alot of experimenting on how to generate audio that feels like the motion of A.I latent step movements, I found that wolfram tones which is an algorithmic music generator process that uses cellular automata and mathematical rules derived from Stephen Wolfram's research in A New Kind of Science.

By assigning specific musical elements such as pitch, rhythm, and instrumentation to the evolving patterns in cellular automata, the wolfram tones output really strange and cool melodies.