Multi-shot Video Generation on 100 hierarchical prompts with transitions
80.95Inter-shot Semantic Consistency ScoreCineTrans-Unet
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| CineTrans-Unet2025.08 | 80.95 | 85.98 | 0.2444 | 72.47 | 0.1457 | 95.98 | 97.25 | 57.47 | 22.24 | |
| CineTrans-DiT2025.08 | 78.58 | 70.03 | 0.1552 | 78.74 | 0.1901 | 96.73 | 97.75 | 65.08 | 21.09 | |
| Wanx2.1-T2V-turbo2025.08 | 64.31 | 23.55 | 0.3002 | 65.16 | 0.2333 | 93.32 | 95.9 | 63.24 | 20.46 | |
| HunyuanVideo2025.08 | 57.23 | 21.11 | 0.4075 | 54.36 | 0.3485 | 94.76 | 96.33 | 60.42 | 20.64 | |
| HunyuanVideo + Cinematron2025.08 | 56.31 | 37.87 | 0.3764 | 60.53 | 0.2855 | 96.06 | 97.21 | 59.78 | 20.82 | |
| StoryDiffusion + CogVideoXI2V2025.08 | 52.14 | — | 0.4966 | 56.6 | 0.3605 | 97.83 | 97.13 | 62.96 | 20.91 | |
| CogVideoX2025.08 | 51.5 | 3.24 | 0.5915 | 62.48 | 0.2226 | 93.1 | 95.82 | 55.09 | 20.61 |