Beyond Reconstruction: Full-Context Generative DiT for Music Generation

Yunjia Li, Menglin Wu, Junyu Dai +13

Hybrid music generators combine the long-range planning of an autoregressive language model with the fidelity of a diffusion- or flow-based acoustic renderer. Yet renderers are tra…