Talking avatar video generation on Long dataset 25 synthesized avatar images, 20s audio clips 1.0
4.01ASEStreamAvatar (baseline)
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| StreamAvatar (baseline)2025.12 | 4.01 | 3.13 | 5.99 | 8.36 | 111.69 | 0.971 | |
| StreamAvatar (+ref sink)2025.12 | 4.64 | 3.64 | 5.82 | 8.27 | 102.16 | 0.973 | |
| OmniAvatarSteps=25, Resolution=480p, Time=36min2025.12 | 4.66 | 3.74 | 6.47 | 8.98 | 15.17 | 0.974 | |
| Hallo3Steps=51, Resolution=480p, Time=32min2025.12 | 4.68 | 3.58 | 4.34 | 9.61 | 59.15 | 0.959 | |
| HY-AvatarSteps=50, Resolution=720p, Time=74min2025.12 | 4.8 | 3.78 | 5.98 | 8.58 | 80.46 | 0.975 | |
| StreamAvatar (+RAPR)2025.12 | 4.84 | 3.81 | 5.99 | 8.18 | 61.07 | 0.961 | |
| StableAvatarSteps=40, Resolution=480p, Time=12min2025.12 | 4.87 | 3.83 | 2.9 | 11.57 | 38.06 | 0.963 | |
| EchoMimicV3Steps=25, Resolution=480p, Time=7min2025.12 | 4.87 | 3.9 | 2.2 | 11.64 | 41.31 | 0.937 | |
| StreamAvatar (+GAN w/o DCA)2025.12 | 4.91 | 3.8 | 6.38 | 8.15 | 29.85 | 0.993 | |
| StreamAvatarSteps=3, Resolution=720p, Time=20s2025.12 | 4.94 | 3.91 | 6.64 | 8.14 | 65.77 | 0.993 |