Video Editing on DAVIS (40 selected object-centric videos)
31.7Prompt Consistency (P.C.)TokenFlow
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| TokenFlowTime (Inference efficiency)=150s2023.12 | 31.7 | 97.7 | 7 | 19.6 | |
| ControlNet per frameTime (Inference efficiency)=50s2023.12 | 31.4 | 91.4 | 36.6 | 58.6 | |
| ControlVideo-edgeTime (Inference efficiency)=120s2023.12 | 31.4 | 97.5 | 6.9 | 23.9 | |
| Text2Video-zeroTime (Inference efficiency)=60s2023.12 | 31.2 | 96.4 | 20.7 | 42.5 | |
| MaskINTTime (Inference efficiency)=22s2023.12 | 31.1 | 95.2 | 9.5 | 27.7 | |
| Tune-a-VideoTime (Inference efficiency)=20min2023.12 | 29.9 | 96.6 | 20.4 | 48.2 |