Video Reconstruction on Wan2.1 Evaluation Set
131Latency (s)XAttn + Ours
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| XAttn + OursBackbone=Wan2.1-1.3B, Resolution=480P2026.05 | 131 | 20.34 | 64.31 | 25.54 | 1.49 | |
| XAttentionBackbone=Wan2.1-1.3B, Resolution=480P2026.05 | 150 | 19.62 | 62.19 | 27.7 | 1.3 | |
| SVG2 + OursBackbone=Wan2.1-1.3B, Resolution=480P2026.05 | 156 | 21.61 | 71.93 | 17.85 | 1.25 | |
| SVG2Backbone=Wan2.1-1.3B, Resolution=480P2026.05 | 170 | 21.56 | 71.89 | 18.43 | 1.15 | |
| DenseBackbone=Wan2.1-1.3B, Resolution=480P2026.05 | 195 | — | — | — | 1 | |
| XAttn + OursBackbone=Wan2.1-14B, Resolution=480P2026.05 | 791 | 20.03 | 70.84 | 23.74 | 1.3 | |
| XAttentionBackbone=Wan2.1-14B, Resolution=480P2026.05 | 843 | 18.67 | 68.01 | 27.61 | 1.22 | |
| SVG2 + OursBackbone=Wan2.1-14B, Resolution=480P2026.05 | 882 | 19.14 | 66.28 | 26.35 | 1.17 | |
| SVG2Backbone=Wan2.1-14B, Resolution=480P2026.05 | 926 | 18.26 | 64.87 | 30.73 | 1.11 | |
| DenseBackbone=Wan2.1-14B, Resolution=480P2026.05 | 1,030 | — | — | — | 1 |