Lecture-Performance
RAVE is a neural synthesis model for real-time performance. It compresses high-dimensional input data into lower-dimensional representations, known as latent spaces, that can be navigated by the performer in visual programming environments through dedicated objects. In this presentation I will reflect on the use of spatialization in a multichannel setup as a way of projecting the model’s latent dimensions into the physical space, with the aim of increasing its interpretability.
In order to bridge RAVE’s internal representations with the outer world, I developed a compositional system based on magnetic scores and controllers that allows the design of complex performative gestures whilst retaining a high degree of independence and control over individual dimensions.
After reporting on my research experience with the Tangible Music Lab’s Dodekaotto spherical setup, I will perform live with the magnetic scores system.