Text-to-Video on MiraData Pexels
99.35SmoothnessMAGI-1
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| MAGI-1Inference steps=20, Chunk size=3 frames2025.12 | 99.35 | 46.88 | 55.69 | 58.43 | |
| SkyReels-V2Inference steps=20, Chunk size=3 frames2025.12 | 99.2 | 44.53 | 55.6 | 57.84 | |
| Self ForcingInference steps=20, Chunk size=3 frames2025.12 | 98.64 | 49.22 | 58.33 | 71.29 | |
| BAggerInference steps=20, Classifier-free guidance scale (w)=6, Time step shift coefficient (k)=8, Chunk size=3 frames, Base model=Wan2.1 1.3B, Sliding window configuration=12 latent frames context2025.12 | 98.61 | 76.56 | 55.35 | 63.41 |