Personalized Human Action Video Generation on MotionX and MotionX++ (val)
589.8FVDLoFA
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| LoFAAdaptation Strategy=Direct Prediction, Conditioning Modality=Text, Backbone=WAN2.1-1.3B2025.12 | 589.8 | 37.19 | 22.83 | |
| LoRAAdaptation Strategy=Per-case Optimization, Backbone=WAN2.1-1.3B2025.12 | 609.5 | 36.62 | 22.69 | |
| LoFAAdaptation Strategy=Direct Prediction, Conditioning Modality=Pose, Backbone=WAN2.1-1.3B2025.12 | 610.7 | 36.87 | 22.97 | |
| Text-to-LoRAAdaptation Strategy=Direct Prediction, Backbone=WAN2.1-1.3B2025.12 | 907.5 | 35.41 | 7.45 |