Audio-driven Digital Human Generation on Short sequence
6.97Sync-CInfiniteTalk
Evaluation results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| InfiniteTalk#Params=14B, FPS=1.312026.02 | 6.97 | 8.65 | 38.56 | 418.43 | 5.07 | 3.97 | 0.91 | 0.9 | 0.96 | |
| EchoTorrent#Params=14B, FPS=10.502026.02 | 6.77 | 8.52 | 38.85 | 415.76 | 5.09 | 3.98 | 0.92 | 0.91 | 0.96 | |
| SoulX-FlashTalk#Params=14B, FPS=8.212026.02 | 6.65 | 8.69 | 38.89 | 418.67 | 5.02 | 3.86 | 0.92 | 0.89 | 0.95 | |
| EchoMimicV3-Flash#Params=1.3B, FPS=2.122026.02 | 6.13 | 9.09 | 42.76 | 496.88 | 4.96 | 3.75 | 0.89 | 0.88 | 0.92 | |
| HunyuanAvatar#Params=14B, FPS=1.312026.02 | 6.12 | 9.11 | 42.54 | 676.28 | 4.96 | 3.67 | 0.89 | 0.85 | 0.89 | |
| FantasyTalk#Params=14B, FPS=1.312026.02 | 4.05 | 11.01 | 45.03 | 603.95 | 4.85 | 3.48 | 0.86 | 0.8 | 0.87 |