CineRecon: Benchmarking Frontier Models for Agentic Cinematic Reconstruction

Key points

• Reconstructing reference videos as editable dynamic 3D scenes is an important path toward controllable video generation for filmmaking: these scenes provide modifiable, renderable visual proxies that enable explicit control over scene layout, subject actions, and camera behavior.1
• Given a complete video, an agent uses tools to construct, animate, render, and refine its reconstruction.1
• This design tests whether agents can reproduce the events in a video when detailed appearance is unnecessary, and how performance changes as representation requirements increase.1
• A vision-language judge compares timestamp-aligned reference and rendered frames across 11 categories, grouped into camera, static, and dynamic consistency dimensions.1

Sources (2)

• [1]CineRecon: Benchmarking Frontier Models for Agentic Cinematic Reconstruction
  Hugging Face Daily Papers · Oct 12, 12:00 AM

  To evaluate agents' ability to reconstruct editable dynamic 3D scenes from dynamic reference videos, we introduce CineRecon, a benchmark comprising 50 reference videos and 360 annotated evaluation clips.

  Reconstructing reference videos as editable dynamic 3D scenes is an important path toward controllable video generation for filmmaking: these scenes provide modifiable, renderable visual proxies that enable explicit control over scene layout, subject actions, and camera behavior.

• [2]CineRecon: Benchmarking Frontier Models for Agentic Cinematic Reconstruction
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 12, 10:57 AM · same content

Extractive summary: sentences quoted from the sources.