DuoTok: Source-Aware Dual-Track Music Tokenization for Vocal-Accompaniment Generation
DuoTok is presented, a source-aware dual-track music tokenizer for vocal-accompaniment generation based on staged disentanglement, suggesting that tokenizer design is a core modeling problem for multi-track music generation, beyond compression alone.
Rui Lin, Zhiyue Wu, Jia-He Lei et al.
· 1 citation