Claude Skill MartBrowse skillsQuick linesLearn by videoTerminal guideWhat is a Skill?
 
← Back to list

video-reference (Finished-Film Breakdown to Production Reference)

Breaks a finished video down into measured cut rhythm, loudness and section structure, exporting a reusable production_reference.json that strips out the source film's names and dialogue.

Image & VideoAdvanced★ 548⑂ 103AI score 8/10Last updated: Oct 3, 2026

What it does

  • Runs a single ffmpeg pass to measure scene cuts, shot-length distribution, cuts per minute, a 10-second cut curve, plus integrated loudness / LRA / true peak and per-second short-term loudness.
  • Collects ambiguous transitions (motion shots, fast push-ins, flashes, dark-scene hard cuts) into review windows and renders frame contact sheets so you can visually confirm and add/remove cuts before any numbers are trusted.
  • Guides labeling of audio ownership (original dialogue vs narration), narrative sections and subtitle style, then strictly separates "facts about this film" from "methods reusable on the next one".
  • Enforces eight validation rules (R1–R8) covering schema closure, time ordering, evidence anchoring and target provenance, and performs a literal leak scan for character names, shared phrases, absolute timecodes and paths before exporting production_reference.json.

Who it's for

  • Creators who produce recap/commentary videos repeatedly and want to absorb a reference film's editing rhythm as numbers, not vibes.
  • Users already running the zenstory video-recap pipeline with its understanding artifacts directory (U).
  • Editors and producers who need the method, not the content, for copyright-safety reasons.

Examples

  1. measure a 5-minute 720p commentary video: get 82 cuts and 11 review windows, then use frames --review to recover 5 missed cuts before any derived metric is used.
  2. Label audio_spans and sections, run check to inspect narration share, per-section cut density and how often audio switches land on picture cuts.
  3. Once error-free, export and drop production_reference.json into the next project's work_dir, recording adopt / adapt / skip per method during scripting.

· · · Install guide · · ·

Try it now, no install

Paste this into Claude to use the skill without installing anything.

Read the instructions in this file and follow them to help me:
https://raw.githubusercontent.com/zenstory-ai/video-recap-skills/HEAD/skills/video-reference/SKILL.md

What I want: (describe your task here)

If Claude can't open the link, open it yourself and paste the contents instead.

↓ If it works for you, download the ZIP below and install it. Then it runs on its own — no pasting each time.

Install in the Claude app (no terminal)
  1. Download the ZIP with the button below.
  2. In Claude, open Settings → Capabilities and turn on 'Code execution and file creation'. (one time)
  3. Go to Customize → Skills → + → 'Upload a skill' and upload the ZIP.
↓ Download ZIP
Install in Claude Code

Let Claude do it — paste this into Claude Code

Install the skill I found on Claude Skill Mart.
Copy the skills/video-reference folder from the GitHub repo zenstory-ai/video-recap-skills into my ~/.claude/skills/video-reference/.
When it's done, tell me in one line what this skill can do.

Install with a command instead

git clone https://github.com/zenstory-ai/video-recap-skills.git && mkdir -p ~/.claude/skills && cp -r video-recap-skills/skills/video-reference ~/.claude/skills/

⚠ This is a third-party skill. Check the source repository before installing.

  1. Open a terminal in your projects folder and clone the repo: git clone https://github.com/zenstory-ai/video-recap-skills.git
  2. Create the skills folder if needed: mkdir -p ~/.claude/skills
  3. Copy the skill in: cp -r video-recap-skills/skills/video-reference ~/.claude/skills/
  4. Verify prerequisites: ffmpeg -version && python3 --version (ffmpeg must include the scdet filter).
  5. Restart Claude Code and invoke it with phrases like "break down this finished video", "production reference", or the Chinese triggers 拆片 / 制作参考.
  6. Point it at a finished video file (mp4/mov/mkv/webm) and, ideally, the understanding-artifact directory U for that same film. It still runs without U, but narration pacing stays empty and the leak scan only sees entities you write yourself.
  7. Keep U/ (the breakdown JSON and reference_frames/) local — it contains source-film facts and stills, so never commit or share it.
View source on GitHub ↗License: MIT