AI Performance Software for Transmediale, Berlin, 2019
Used for the two Highlight Performances at Transmediale 2019 in Berlin.
Using techniques from Nick Collins’s AIrtist project, music from Actress’s back catalog was analysed for chord progressions and melody. The melodies were modelled using Magenta RNN networks and a novel graphic score interface was built on top to generate seed content and sequence the generation of melodic motifs.
Actress used this software with Ableton in a live collaboration with his Young Paint alter ego, with live visuals by Werkflow.
Actress was after something that was took inspiration from both his previous work and his graphic scores. We also wanted something that could be played and explored, as opposed to something entirely autonomous. He’s a great musician and the idea was to help him explore new ideas and break out of the affordances of his keyboards and DAWs.
Obviously there is a market for entirely AI produced music that pumps out end products, but once you have a black box that can generate music, as far as I’m concerned that’s not where the journey ends. A lot times things make fun demos, but then fail to make it into the workflows of musicians, even less musicians who don’t code, even though computers may play a large part in their practice. Given that we can take a past collection of melodies and make a model that generates novel content, this project was about exploring how you can package this tech into something musicians can use like they would any other instrument.