Search papers, labs, and topics across Lattice.
1
0
3
2
Compressing audio semantics into just 128 dimensions doesn't just reduce DiT modeling burden; it actually *improves* audio generation quality across diverse domains.
