ParamFlow: A generative framework for augmented control over audio tools

Unfortunately this talk is only open to CHIA and Music Faculty members.

Synthesizers, audio effects, and audio workstations are as much a part of our musical heritage as the music made with them. In developing the next generation of AI-enhanced music technology, I argue that rather than seeking to replace these tools and erase this heritage, we should embrace them and build our technology around them. To this end, I present ParamFlow: a generative framework that acts as a translation layer between audio and tool parameters, enabling augmented and permeable control over the music studio. In this talk, I introduce the technical foundations of ParamFlow and demonstrate its application to synthesizer sound matching and audio effects estimation.

Ben Hayes is a research scientist at Sony Computer Science Laboratories – Paris, where he works as part of the Music Team on tools for audio and music production. He conducted his PhD research at Queen Mary University of London’s Centre for Digital Music, where he focused on incorporating digital signal processing techniques into deep neural networks for audio generation and processing. Previously, he was Music Lead at the pioneering generative music startup Jukedeck, and completed Research Scientist internships at Spotify, Sony CSL, and Bytedance. Before transitioning to computer science, he worked as a music producer and toured internationally as a musician.