Chong Jing

Continuous autoregressive models operate directly on continuous representations and Bottlenecked Representation Alignment and Modality--Time RoPE are introduced to strengthen musical semantic structure in Composer hidden states and temporal alignments across modalities.

P-MUSE achieves unprecedented flexibility in music synthesis by allowing users to generate and edit music with or without aligned MIDI inputs.

Anysynth achieves superior zero-shot instrument cloning by dynamically retrieving acoustic details, outperforming traditional embedding-based methods in timbre fidelity and audio quality.