Audio-driven Generation on TalkBench Long (> 5 min) 1.0 (test)
3.15ASELiveAvatar
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| LiveAvatarReal-time inference support=true2025.12 | 3.15 | 3.06 | 0.96 | 12.73 | 98.19 | 97.39 | 99.26 | 98.87 | 20.88 | |
| InfiniteTalkLoRA version=true2025.12 | 3.12 | 3.03 | 1.51 | 12.28 | 98.97 | 97.32 | 99.59 | 99.5 | 5.1 | |
| SoulX-FlashTalkReal-time inference support=true2025.12 | 3.12 | 3.04 | 1.61 | 12.25 | 99.29 | 98.36 | 99.63 | 99.58 | 32 | |
| DittoReal-time inference support=true2025.12 | 3.11 | 3.03 | 0.83 | 13.45 | 99.8 | 99.23 | 99.76 | 99.89 | 0.53 | |
| StableAvatar2025.12 | 3.11 | 3.02 | 0.7 | 12.73 | 98.38 | 96.42 | 99.42 | 98.99 | 0.64 | |
| OmniAvatar2025.12 | 2.92 | 2.82 | 0.99 | 12.8 | 85.42 | 86.59 | 99.61 | 99.57 | 0.16 |