Audio-driven portrait video generation on HDTF (test)
9.489SCVASA-1
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| VASA-1Resolution=512 x 512, Hardware=RTX 4090 GPU2026.06 | 9.489 | 5.928 | 74.2 | 114.863 | 37.9 | |
| Ours M=1Resolution=512 x 512, M (Number of reference images)=1, Hardware=RTX 4090 GPU2026.06 | 9.196 | 6.024 | 70.7 | 84.569 | 23.3 | |
| OursM (number of reference images)=2, Resolution=512x512, Hardware=Single H100 GPU2026.06 | 9.056 | 6.175 | 73.9 | 49.4 | 42.3 | |
| OursM (number of reference images)=3, Resolution=512x512, Hardware=Single H100 GPU2026.06 | 8.998 | 6.226 | 73.9 | 43.27 | 42.3 | |
| OursM (number of reference images)=1, Resolution=512x512, Hardware=Single H100 GPU2026.06 | 8.943 | 6.286 | 69.9 | 62.3 | 42.3 | |
| SonicM (number of reference images)=1, Resolution=512x512, Hardware=Single H100 GPU2026.06 | 8.799 | 6.602 | 68.9 | 43.92 | 1.7 | |
| Hallo2M (number of reference images)=1, Resolution=512x512, Hardware=Single H100 GPU2026.06 | 7.547 | 7.819 | 24.7 | 88.17 | 1.2 | |
| SadTalkerResolution=512 x 512, Hardware=RTX 4090 GPU2026.06 | 7.497 | 7.49 | 47.9 | 193.843 | 3.8 | |
| HalloM (number of reference images)=1, Resolution=512x512, Hardware=Single H100 GPU2026.06 | 7.457 | 7.841 | 24.2 | 90.88 | 1.2 | |
| Hallo3M (number of reference images)=1, Resolution=512x512, Hardware=Single H100 GPU2026.06 | 7.256 | 8.596 | 33.7 | 76.43 | 0.27 | |
| EchoMIMIC-DistilledM (number of reference images)=1, Resolution=512x512, Hardware=Single H100 GPU2026.06 | 5.513 | 9.35 | 34.8 | 174.061 | 13.3 | |
| EchoMIMICM (number of reference images)=1, Resolution=512x512, Hardware=Single H100 GPU2026.06 | 5.291 | 9.557 | 34.1 | 143.62 | 1.4 | |
| FantasyTalkingM (number of reference images)=1, Resolution=512x512, Hardware=Single H100 GPU2026.06 | 4.167 | 11.144 | 40.7 | 89.726 | 0.36 |