loaf-ai was built for the BitRate: Machine Learning & Music hackathon sponsored by Gray Area and Google Magenta.

loaf-ai allows the listener to assemble a lo fi hip hop track using piano, drum, and sound effect loops. While the tracks are loaded into Tone.js, Magenta's Music RNN generates 40 samples of acoustic guitar improvisations to play over the selected chord progressions.

The resulting composition is a 15-minute-long jam to study / work / relax / to that is complex enough to be musically interesting, but not so unpredictable as to be a distraction. To add some visual interest and draw the listener's attention to the AI soloist's contributions, the guitar track is highlighted as a waveform at the bottom of the playing browser, adapted from Jason Sigal's p5-music-viz.

Read more -> https://kathrynisabelle.com/loaf-ai-writeup

Play -> https://kathrynisabelle.com/loaf-ai (probably only works in Chrome, sorry)