GENERATIVE SINGING STYLE TRANSFER ACROSS GENRES

Saanvi Bhargava (The Harker School), Ethan Chu (Monta Vista High School), Matthew Lee (Riverdale Country School), Chuyang Chen (New York University), Kelvin Walls (New York University), Bea Steers (New York University), Iran R Roman (Stanford University)*

lbd-470-generative-singing-style

This paper will be presented in person

Abstract:

Voice style transfer has focused on transforming speech between speakers, leaving singing style, independent of speaker, unexplored. We introduce SingStyleTransfer, a VAE-GAN to perform singing style transfer across genres. The model is evaluated on the SingStyle111 dataset for its ability to carry out genre-to-genre transformations.