Our project focuses on leveraging the timbral characteristics of audio signals to manipulate parameters in real time within a sound generation tool. We've designed an interface that allows users to visualize various timbre-related features of an audio input, empowering them to select which aspects of timbre they wish to explore further. Through this interface, users can seamlessly map these chosen timbre features to synthesis parameters.
As a proof of concept, we've demonstrated the efficacy of this mapping by controlling the latent space of pre-trained RAVE models. Additionally, we've provided all the essential components for integrating a VST synth plugin, further enhancing the versatility and applicability of our approach.
To infinity and beyond - explore, navigate, generate, timbre space.
Arjun Bahuguna, Andrea Bolzoni, Mihai Traista & Francesco Di Maggio