Audio-driven portrait video generation on PortraitOneMin (test)
42.3FPSOurs
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| OursM (number of reference images)=1, Resolution=512x512, Hardware=Single H100 GPU2026.06 | 42.3 | 8.537 | 6.619 | 0.648 | 91.964 | |
| OursM (number of reference images)=2, Resolution=512x512, Hardware=Single H100 GPU2026.06 | 42.3 | 8.438 | 6.688 | 0.677 | 81.517 | |
| OursM (number of reference images)=3, Resolution=512x512, Hardware=Single H100 GPU2026.06 | 42.3 | 8.546 | 6.648 | 0.659 | 73.693 | |
| EchoMIMIC-DistilledM (number of reference images)=1, Resolution=512x512, Hardware=Single H100 GPU2026.06 | 13.3 | 5.531 | 9.187 | 0.201 | 201.899 | |
| SonicM (number of reference images)=1, Resolution=512x512, Hardware=Single H100 GPU2026.06 | 1.7 | 8.185 | 7.031 | 0.598 | 95.047 | |
| EchoMIMICM (number of reference images)=1, Resolution=512x512, Hardware=Single H100 GPU2026.06 | 1.4 | 4.863 | 9.594 | 0.208 | 177.141 | |
| HalloM (number of reference images)=1, Resolution=512x512, Hardware=Single H100 GPU2026.06 | 1.2 | 6.721 | 8.275 | 0.21 | 179.072 | |
| Hallo2M (number of reference images)=1, Resolution=512x512, Hardware=Single H100 GPU2026.06 | 1.2 | 6.829 | 8.244 | 0.196 | 162.672 | |
| FantasyTalkingM (number of reference images)=1, Resolution=512x512, Hardware=Single H100 GPU2026.06 | 0.36 | — | — | — | — | |
| Hallo3M (number of reference images)=1, Resolution=512x512, Hardware=Single H100 GPU2026.06 | 0.27 | 6.836 | 8.724 | 0.301 | 175.34 |