notesum.ai
Published at November 25DreamRunner: Fine-Grained Storytelling Video Generation with Retrieval-Augmented Motion Adaptation
cs.CV
cs.AI
cs.CL
Released Date: November 25, 2024
Authors: Zun Wang1, Jialu Li1, Han Lin1, Jaehong Yoon1, Mohit Bansal1
Aff.: 1UNC Chapel Hill
![[Uncaptioned image]](https://arxiv.org/html/2411.16657v1/x1.png)
| Method | Image | Fine-Grained Text | Full Text | Transition | |||
| CLIP | DINO | CLIP | ViCLIP | CLIP | ViCLIP | DINO | |
| VideoDirectorGPT [33] | 54.3 | 9.5 | 23.7 | 21.7 | 22.4 | 22.5 | 63.5 |
| VLogger [74] | 62.5 | 41.3 | 23.5 | 23.1 | 22.5 | 22.2 | 73.6 |
| DreamRunner (Ours) | 70.7 (+13.1%) | 55.1 (+33.4%) | 24.7 (+5.11%) | 23.7 (+2.60%) | 24.2 (+7.56%) | 24.1 (+8.56%) | 93.6 (+27.2%) |