Preprint
Sep 2026
DiffSynth-Music: Audio-Conditioned KV-Cache Adapters for Controllable Music Generation
DiffSynth-Music is introduced, a framework that adds composable audio conditioning to a music synthesis backbone through layer-wise key-value injection and releases the three template models, Control, Prosody, and Reference, to support research and creative applications in controllable music generation.
Zhong-Jie Duan, Sheng-Chuan Gao, Hong Zhang et al.
· 0 citations