For years, Lee and I have been collaborating on artificial singing voices for his music productions. Following performances at Ina GRM and Unsound, Lee's album Models will be releasing on Hyperdub in October 2023.
Over the course of our work I have advised on machine learning concepts, researched state-of-the-art models for singing voice synthesis, collated datasets, trained models, and maintained Colab notebook for Lee to generate singing voices.
A playlist of bootlegs I've made using some source separation tools and my autoencoder effects.
Spleeter was used to separate tracks into bass, drums, vocals, and "other" stems. I separately passed them through ANNe to edit their spectral content.