Video-conditioned sound-speech joint generation on V2C (test)
19.4WERVSSFlow-M
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| VSSFlow-MModel size=Medium2025.09 | 19.4 | 24.7 | 2.07 | 2 | 6.83 | 177.16 | 0.91 | 1.07 | |
| LoVA | SpeakerConditioning=Speaker2025.09 | 25.98 | 20.02 | 1.33 | 1.28 | 8.68 | 285.84 | 0.97 | 1.27 | |
| LoVA+StyleConditioning=Style2025.09 | 37.56 | 19.58 | 1.3 | 1.27 | 9.28 | 296.64 | 1.05 | 1.31 | |
| MMAudio+SpeakerConditioning=Speaker2025.09 | 41.47 | 18.39 | 1.32 | 1.91 | 4.32 | 229.65 | 0.89 | 1.14 | |
| MMAudio+StyleConditioning=Style2025.09 | 56.86 | 17.78 | 1.3 | 1.92 | 4.48 | 240.7 | 0.91 | 1.16 |