Search papers, labs, and topics across Lattice.
2
0
3
Natural-language instructions can now handle both zero-shot speech synthesis and surgical acoustic editing within a single unified model, operating at 4-step distilled inference speeds without classifier-free guidance.
FullDiT not only outperforms leading commercial music generators but also redefines how we approach music rendering by leveraging full-context generation techniques.
