Junan Zhang

Reward-based fine-tuning of a discrete diffusion model significantly enhances synthesizer inversion performance, outperforming traditional methods in audio matching tasks.

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.