Multi-shot video storytelling on ST-Bench
64.52Aesthetic QualityWan2.2-T2V
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Wan2.2-T2VDescription=pretrained base model independently generating shots2025.12 | 64.52 | 21.74 | 24.52 | 39.37 | 42.48 | |
| StoryMemBase Model=Wan2.2-I2V-A14B, LoRA Rank=128, Memory Sink Size=32025.12 | 61.33 | 22.89 | 23.13 | 50.65 | 53.37 | |
| StoryDiffusion + Wan2.2-I2VBase Model=Wan2.2-I2V2025.12 | 60.85 | 22.88 | 23.49 | 46 | 47.32 | |
| IC-LoRA + Wan2.2-I2VBase Model=Wan2.2-I2V2025.12 | 57.04 | 21.31 | 21.81 | 41.1 | 42.96 | |
| HoloCine2025.12 | 56.53 | 21.99 | 21.25 | 46.28 | 50.05 |