My friend Josh has been working on a small language model that predicts C64 music, and he’s been using some of my music to test it out.
It’s called preframr – Predicting Register Encoded Frames Rendered As Musical Results – and it’s available as open source on GitHub. You can feed it some C64-tunes, train the model, and generate new C64-music.
There’s still work to be done, but it’s starting to output things that are 1) not too similar to the original and 2) makes sense as music. :D