#inktober: an experiment


    
Day 18: Frankensheriff

This one's a weird one. It also uses a Magenta music model called Music VAE. One of the things this model can do is encode note sequences in a lower dimension space, and then decode points from that space into a new note sequence. Normally, if you encode a note sequence and then decode that encoding, you get the original one. Here, I am encoding the drum beat that represents the sheriff (see day 16 for how that works), adding a little bit of noise, and then decoding the noisy encoding. What you get is...this.