VLA + MLMLMLM

Lucy Strauss, Goldsmiths, University of London

Artwork Description:

VLA + MLMLMLM is a live performance for violist and MLMLMLM, a bespoke deep generative model that reconceives listening as an embodied, multisensorial process. The work employs generative modeling as a speculative instrument to ask: what aesthetic formations emerge when listening is routed through musculature rather than auditory perception? MLMLMLM is trained to translate electromyographic muscle signals produced during viola performance into a latent representation of viola timbre, enabling the agent to predict and synthesise viola-like sound exclusively from bioelectric bodily data. In performance the acoustic viola remains audible while MLMLMLM constructs a parallel sonic response derived solely from live muscle signals and the model’s learned latent space. The piece stages an encounter between human corporeality and statistical probability, foregrounding experiential dimensions of instrumental practice typically hidden from audiences. It surfaces emergent aesthetics that arise when an artificial listener operates from within the performer’s body, reframing listening encounters of live performance. VLA + MLMLMLM has been presented in Cape Town (ZA), Gothenburg (SE), and Delft (NL) and continues to evolve through ongoing artistic and technical research. Muscle Listening Machine Learning Model for Live Music * MLMLMLM is trained on an embodied archive of audio and EMG signals collected from the artist’s body.

Project Collaborators:

This work forms part of a PhD funded through a Consortium for the Humanities & the Arts South-East England AHRC Doctoral Training Partnership. The custom dataset used for MLMLMLM training was recorded during a week-long studio residency at the Studios for Electroacoustic Music, Akademie der Kuenste, Berlin.