Text-to-Video Generation on LTX-Video-2B 4x native length (481 frames) (dev)
0.9726Subject ConsistencyDirect
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| DirectInference Time=∼13 mins2026.05 | 0.9726 | 0.9682 | 0.98 | 0.1933 | 0.5448 | 0.6468 | |
| FreeLongInference Time=∼22 mins2026.05 | 0.9699 | 0.964 | 0.9804 | 0.0311 | 0.543 | 0.6478 | |
| FreeSpecInference Time=∼22 mins2026.05 | 0.9622 | 0.9677 | 0.9892 | 0.2478 | 0.5464 | 0.6382 | |
| FreeNoiseInference Time=∼12 mins2026.05 | 0.9618 | 0.9614 | 0.9822 | 0.0556 | 0.523 | 0.6528 | |
| FreePCAInference Time=∼21 mins2026.05 | 0.961 | 0.9637 | 0.9796 | 0.2111 | 0.5453 | 0.6332 | |
| Sliding WindowInference Time=∼12 mins2026.05 | 0.9559 | 0.962 | 0.9795 | 0.2311 | 0.5442 | 0.6282 |