![[OLD FALL 2017] 15-104 • Introduction to Computing for Creative Practice](https://courses.ideate.cmu.edu/15-104/f2017/wp-content/uploads/2020/08/stop-banner.png)
Project: Computer Replicating Voices
Video:
Key times: 9:30 -> 11:48
Artist (?): Cary Huang
Year: 2017
Although this “project” is from a content creator on Youtube, I thought it was fascinating how a programmer could create a specific algorithm to teach a computer to learn a specific voice. This was all done through the program, ARSS, or The Analysis & Resynthesis Sound Spectograph, Python, and another program called HyperGAN. The ARSS transfer sounds into images which is also modified through Python, then the HyperGAN software uses Convolutional and Deconvolutional neural networks to generate new images in the same style. The ARSS then transfer the images back into audio to become audible.
I find it amazing how not only can we start to teach computers to slowly learn through parameters, but also that such complex process and discovery was put into a YouTube video as if it were “just something cool”. Technology has come far enough to dismiss this amazing demonstration as “a cool video” that is shared around.
This video didn’t highlight an “artist”, rather it showed how complex computers can develop its neural networks. This creator also showed how AI could learn to play Jazz or Baroque music through hours of “training”, which brings another question… is a completely computer generated (through human made parameters) art? I think the thought and artistry behind the codes and programs behind such acts allow AI creation to be artworks.