Image-to-Video Generation on HumanVid curated 500 real-world videos (evaluation set)
0.42LPIPSTokenMotion-I
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| TokenMotion-Ibackbone=CogVideoX-2B2025.04 | 0.42 | 74.65 | 332.27 | 3.77 | 0.22 | 2.98 | 83.82 | 3.63 | |
| ImageConductorimplementation=cascaded joint-control2025.04 | 0.43 | 131.36 | 878.59 | 4.32 | 0.44 | 5.44 | 164.28 | 9.18 |