Subject and Motion Customization on Human Evaluation 50 groups: 5 motion patterns and 10 subjects
82.8Text AlignmentDreamVideo
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| DreamVideoComparison Baseline=LoRA2023.12 | 82.8 | 53.5 | 83.7 | 67.4 | |
| DreamVideoComparison Baseline=AnimateDiff [21]2023.12 | 72.3 | 62.3 | 82.4 | 64.2 | |
| DreamVideoComparison Baseline=ModelScopeT2V [64]2023.12 | 62.4 | 66.2 | 56.6 | 51.5 | |
| ModelScopeT2VComparison Target=DreamVideo2023.12 | 37.6 | 33.8 | 43.4 | 48.5 | |
| AnimateDiffComparison Target=DreamVideo2023.12 | 27.7 | 37.7 | 17.6 | 35.8 | |
| LoRAComparison Target=DreamVideo2023.12 | 17.2 | 46.5 | 16.3 | 32.6 |