Search papers, labs, and topics across Lattice.
Shanghai Jiaotong University
1
0
3
Current video-to-audio models are surprisingly bad at generating speech and singing, exposing critical gaps in multimodal understanding.
