Melody Mixer

March 2018 | By Torin Blankensmith, Kyle Phillips

A fun way to explore melodies using machine learning


Melody Mixer is a simple demo that shows what you can do by combining machine learning tools like TensorFlow.js and MusicVAE.js with easy-to-use web tools like P5.js and Tone.js. It explores a simple idea: what would it sound like to blend between two different musical melodies?

It uses MusicVAE.js, a framework from the Magenta team at Google, to blend between the melodies. This all happens in realtime, locally in your browser, using TensorFlow.js.

Built by Torin Blankensmith and Kyle Phillips at Google Creative Lab in collaboration with Adam Roberts from the Google Magenta team. To learn more about how Melody Mixer was made and start making your own projects, check out this blog post.