Rima Glottidis: Experimenting Generative Raw Audio Synthesis for a Sound Installation

Biologically-inspired algorithms such as artificial neural networks have been used extensively by computer music researchers, for generative music and algorithmiccomposition. The recent introduction of raw audio Machine Learning (ML) techniques, however, represents a significant leap because they seem to be able to learnboth high-level (event) and low-level (timbre) information at once. Employing such techniques for creative purposes is very challenging at this early stage sincethere is lack of method, experience, and their computational cost is very high. In this paper, we describe the technical and creative process behindRima Glottidis,an installation based on material from the homonym musical work from the artistØkapi,processedusingrawaudioMLtechniquesforaninstallationpremieredatBlooming festival, Italy. Technical issues are described and critical aspects are reported as a reference for future projects.