Abstract 23

One in a series of low-fi glitches reimagined by a neural network. Each piece was progressively grown from a 32x32 pixel starter using a network trained to upscale, or `up-imagine`, images.