Talking Avatar Generation on long-form videos (test)
56.22FIDAvatarForcing
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| AvatarForcingLatency (s/frame)=0.0342026.03 | 56.22 | 737.92 | 91 | 5.64 | 9.26 | |
| LiveAvatar (5 GPUs)Latency (s/frame)=0.0472026.03 | 57.81 | 765.64 | 90 | 5.46 | 9.12 | |
| StableAvatarLatency (s/frame)=7.402026.03 | 58.8 | 739.72 | 90 | 2.35 | 11.38 | |
| HunyuanAvatarLatency (s/frame)=14.402026.03 | 76.58 | 870.16 | 91 | 2.71 | 9.34 | |
| OmniAvatarLatency (s/frame)=6.212026.03 | 81.12 | 1,430.79 | 86 | 4.21 | 10.82 | |
| WanS2VLatency (s/frame)=16.002026.03 | 88.73 | 870.34 | 74 | 5.13 | 10.36 | |
| MultiTalkLatency (s/frame)=3.752026.03 | 106.04 | 1,788.19 | 76 | 5.26 | 9.39 | |
| Hallo3Latency (s/frame)=9.602026.03 | 122.64 | 1,371.98 | 83 | 3.9 | 10.42 | |
| FantasyTalkingLatency (s/frame)=4.202026.03 | 144.71 | 1,649.23 | 76 | 4.3 | 13.18 | |
| EchoMimicLatency (s/frame)=2.252026.03 | 147.13 | 2,102.42 | 79 | 3.44 | 10.91 |