“To a certain extent, the methodology is fairly straightforward. You take a lot of audio from a speaker. You then cut that up into tiny little pieces. Each piece is a little sound. … In order to then produce a new sentence, you then take those sounds, you rearrange them, and you stick them back together again.”