Text-to-Video Generation on Wan2.1-1.3B (4x native length)
0.9865Subject ConsistencyDirect
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| DirectInference Time=∼38 mins2026.05 | 0.9865 | 0.9752 | 0.9876 | 0.0468 | 0.5858 | 0.6274 | |
| FreeSpecInference Time=∼56 mins2026.05 | 0.9863 | 0.9745 | 0.9902 | 0.32 | 0.6104 | 0.6415 | |
| FreePCAInference Time=∼55 mins2026.05 | 0.9857 | 0.9734 | 0.9853 | 0.2177 | 0.6005 | 0.6372 | |
| FreeLongInference Time=∼61 mins2026.05 | 0.9849 | 0.9731 | 0.9856 | 0.2514 | 0.6057 | 0.6374 | |
| Sliding WindowInference Time=∼25 mins2026.05 | 0.9791 | 0.9677 | 0.9887 | 0.3955 | 0.6018 | 0.6424 | |
| FreeNoiseInference Time=∼25 mins2026.05 | 0.9777 | 0.9667 | 0.9847 | 0.2395 | 0.6074 | 0.6612 |