Hear What You See: Video-to-Audio Generation with Diffusion Transformer and Semantic-Temporal Alignment-Ranked Direct Preference Optimization
Open paper
Benchmark
Model
Rank
Results
video-to-sound-generation-on-vggsound-v2a
VisioSonic w/ STAR-DPO (44.1kHz)
–
FD (PaSST): 55.48
KL (PaSST): 1.29
ImageBind Score: 33.1
…
Rank counts only results with a code link.