Talking Face Generation on User Study (test)
7.2Lip-sync AccuracySynctalk
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| SynctalkMethod Variant=origin, Training Time=3h 10min2025.11 | 7.2 | 5.84 | 7.41 | 6.35 | 0.14 | |
| SynctalkMethod Variant=ours 5s, Training Time=29min2025.11 | 6.59 | 5.55 | 6.28 | 5.64 | 0.83 | |
| TalkingGaussianTraining Time=2h 8min2025.11 | 6.12 | 6.06 | 6.1 | 5.88 | 0.19 | |
| GaussianTalkerTraining Time=4h 4min2025.11 | 5.75 | 5.63 | 5.81 | 5.48 | 0.09 | |
| InstagMethod Variant=ours 5s, Training Time=24min2025.11 | 5.59 | 5.11 | 5.61 | 5.22 | 0.9 | |
| ER-NeRFTraining Time=3h 52min2025.11 | 5.57 | 5.24 | 5.24 | 5.11 | 0.09 | |
| RAD-NeRFTraining Time=4h 20min2025.11 | 5.53 | 5.55 | 6.26 | 5.66 | 0.08 | |
| InstagMethod Variant=origin, Training Time=23min2025.11 | 5.01 | 4.39 | 5.41 | 4.48 | 0.8 |