eternityspring/reelbench-skills

What it solves

It provides a set of specialized tools (skills) for AI coding assistants like Claude Code and Codex to perform professional video analysis and synthesis. It automates the tedious process of "shot-listing" (breaking down a finished video into a detailed analysis table) and creating synchronized video-information overlays.

How it works

The project implements two primary skills:

• video-shots: Uses ffmpeg to measure cut points and durations, then leverages an AI model to analyze shot types, camera movement, and rhythm. It includes 15 quality gates to ensure accuracy.
• video-sync: Synthesizes a video that pairs the original footage with its corresponding shot information, ensuring the text updates and highlights in sync with the video playback. Layouts can be customized via CSS.

Who it’s for

Video editors, directors, and AI researchers who want to automate the breakdown of video content into structured data for analysis or presentation.

Highlights

• Integration: Designed as plugins for Claude Code and Codex.
• Zero Dependencies: Requires only Node.js and FFmpeg; no external npm packages or API keys are needed.
• Interactive Reports: Generates single-file HTML reports with embedded players, synchronized highlighting, and searchable shot tables.
• Detailed Analysis: Captures shot duration, scale, category, camera movement, and rhythm, supported by frame-difference motion curves.