Neuracappella (2023) is an interactive audio A.I. experiment to explore new kind of interaction with machine learning models. We trained an RAVE model on 48+ hours of singing voice data, and run real-time inferences in browser with WebGL2. Users can navigate the latent space by drawing lines on the screen, the model will then follow these tracks to generate corresponding singing voices in real-time, creating a nonlinear audiovisual experience.
