Michaelqaz/SCoPE-Gallery

SCoPE Project Gallery

This dataset contains the 182 generated result videos used by the project page for SCoPE: Sightline-Coordinate Positional Encoding for Video Diffusion Transformers. Each clip is generated from a single input frame and a target camera trajectory. The picture-in-picture overlay visualizes the commanded camera motion.

Links

• —Project page
• —Paper
• —Code
• —Model

Structure

The MP4 files remain individually addressable so the project page can stream one video at a time with standard browser range requests.

videos/
  <gallery-video>.mp4          # original quality, loaded by the lightbox
videos_crf22/
  <gallery-video>.mp4          # CRF 22 web preview, loaded by gallery cards
metadata.csv

The two video directories use identical filenames. The project page streams the smaller CRF 22 version while browsing and requests the corresponding original only when the viewer opens the lightbox.

metadata.csv contains:

• —file_name: the path to the MP4 file
• —section: the project-page section
• —gallery: the gallery/deck label
• —motion: the displayed camera-motion label

Intended use

These videos are research-result examples for studying controllable camera motion in generated video and for reproducing the SCoPE project-page gallery. They are not training data for the released SCoPE model.

License

The gallery videos and metadata are released under CC BY-NC 4.0.

Citation

@article{yin2026scope,
  title={SCoPE: Sightline-Coordinate Positional Encoding for Video Diffusion Transformers},
  author={Yin, Minghao and Lu, Jiahao and Hu, Wenbo and Zhao, Wang and Shan, Ying and Han, Kai},
  journal={arXiv preprint arXiv:2606.27345},
  year={2026}
}