Ghost in the Piano is a project I have developed during my PhD practice which explores the intersection between AI, musical improvisation and robotics. The project has taken various forms, from an installation exhibited in the Science and Media museum, https://photos.app.goo.gl/8hyXLcyNrZC67AUs8 to a complete musical performance system which I have used to perform live improvised music https://drive.google.com/file/d/11pNACsryKmjx7VgLd_15VrIx7LeLd6Jh/view?usp=drive_link . The latest phase of the project is to use the robotic piano I have built as part of an AI controlled autonomous music system. The piano was initially influenced by the work of artists such as Aphex Twin, Sarah Nicholls and Felix’s Machines.
The main instrument is a modified upright parlour piano augmented with eight motors, eight solenoids, two actuators and two loudspeakers.
The motors rotate various chains and strikers across the strings generating acoustic sound. The solenoids create percussive sounds against the sound board whilst the two actuators and loudspeakers resonate the strings with a synthesis engine running in Max Msp.
The listening system uses the FluCoMa system developed by the University of Huddersfield to actively listen to a performing guitarist. In particular it uses the fluid.mfcc object to listen to incoming audio from the guitar and catogorise the playing technique . Depending on what technique is being used (e.g. harmonics) the AI influences the output of a wavetable based sequencer and synthesizer. This creates a generative accompaniment to the guitarist with acoustic sound generated by the robotic piano and synthesized sounds that resonate the piano strings and interact in turn with the acoustic sounds.
The system explores the idea of an AI based system as both accompanist and equal partner in an improvised musical performance. The patterns generated by the system influence what the guitarist plays and conversely the guitarist has an influence over the AI system. The piece explores this relationship with parts where the guitarist obviously leads the improvisation before the AI takes over. I am interested in exploring the tensions in AI influenced systems, where the AI might do something unexpected and how that relates back to musical improvisation.
Figure 1. Interface for the Wavetable based sequencer / synthesiser
I would like Ghost In the Piano to be performed in Wolfson College Auditorium. I perform with the piano playing guitar.
Technical/Stage Requirements
• Stage area for upright piano and large table 2 x 4 m table. Stereo PA with Subwoofer.
Access to auditorium via lift as piano is very heavy on wheels and will fit in standard lift – go through standard doors etc. Help with load out and load in - ideally 3 people.
Outputs for the Ghost in the Piano system are :
Stereo out from laptop to DI box (please supply DI)
2 stereo microphones ( good condenser mics on stands) to pick up acoustic and synthesised sounds from piano.
4 power sockets for Piano and computers.
Program Notes
Ghost in the Piano is a project I have developed during my PhD practice which explores the intersection between AI, musical improvisation and robotics. The project has taken various forms, from an installation exhibited in the Science and Media museum, to a complete musical performance system which I have used to perform live improvised music. The latest phase of the project is to use the robotic piano I have built as part of an AI controlled autonomous music system. The work was initially influenced by the work of artists such as Aphex Twin, Sarah Nicholls and Felix’s Machines.
Ghost in the Piano has been made mostly of reclaimed materials. Namely the Parlour piano that was donated to the project by Jethro Bagust. The wood, plastic, power supplies and cabling has mostly been reclaimed. The carbon footprint of the project has been very low with transport being the largest overhead. The code when in a suitable state will be shared on Github as part of my PhD thesis.
Acknowledgements
The author would like to thank the support of my late father John Eyes, my PhD supervisor and my family and friends.
This work has been supported by Annie Jameson and the National Science and Media Museum.
Creative Commons Attribution 4.0 International License (CC-BY 4.0)
