M12: ‘b.io’: an Audio-Visual Miniature for Saxophone and Computer

Tenor saxophonist and researcher Mark Hanslip emerged in the mid-2000s as a key player on the London jazz scene, gigging nationally and internationally with groups including Outhouse, Nostalgia 77, Jonathan Bratoeff Quartet and Twelves, and has appeared on over 30 recordings on labels including Babel, F-iRE, Tru Thoughts, FMR and Tombed Visions. Since relocating to the north of England, he has co-led organ trio The Revival Room with keyboardist Adam Fairhall, played in trio with Federico Reuben and Paul Hession, toured with his improvising group HTrio plus guest US trumpeter Nate Wooley and performed with guitarist/composer Elliot Sharp and saxophonists Evan Parker and Paul Dunmall. His doctoral practice-led research at the University of York examines the applications of machine learning to the systematic processes and creative outcomes within improvised music.

‘b.io’ is an output of my research into applications of generative machine learning to the practice of free improvisation. This practice-based research is anchored to my longstanding creative practice as an improvising saxophonist, while the presentation style reflects a recent fascination with visualisation of audio using deep learning models. The piece represents the outcome of a process of recording audio datasets, training SampleRNN models of the data, generating and curating samples from them, interacting with them in real-time and finally visualizing the recorded outputs. By making this work I hope to signpost the usefulness and fun of generative machine learning models of raw audio for practitioners of improvised music.