Latest / Kana & Mari's SoundRepos (English)

WelkinYang/Learn2Sing2.0
Learn2Sing 2.0 is the official demo and presentation repository for a target-speaker singing voice synthesis system that generates singing for speakers without singing data. The documented approach combines pitch-conditioned acoustic modeling, a diffusion decoder with fast sampling, and mutual-information-based disentanglement of speaker and style embeddings; the included website presents audio demonstrations for teacher and student speakers.
The skinny
The skinny isn't ready yet — notes appear once the transcript is processed.