Reproduce “VIVID: A Visual Harness for Agentic Video Editing” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Wenhan Yu, Zhaoxi Zhang, Zhilong Tan, Weikang Li, Lei Sha
VIVID equips coding agents to edit video by adding on-demand audiovisual inspection, persistent evidence linked to media versions and timestamps, and review of rendered outputs, while retaining native tools. On AgenticVBench, GPT-5.6-sol with Codex scores 58.6% versus 38.4% for native Codex, a 20.2-point gain; this supports improvement for that model–harness pair, not video editing generally.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.