AI Inspiration WorldAdded Nov 4, 2026
ffmpeg-skill
A Skill that gives agents such as Claude Code, Cursor and Codex 42 local FFmpeg editing tools, with no cloud needed.
- Ecosystem
- Agent Skills
- Skill category
- Media
- Creator / organization
- kajisho5
What it is
ffmpeg-skill is an Agent Skill by GitHub user kajisho5 for Claude Code, Cursor, Codex and any agent that reads SKILL.md, and it can also be used as MCP tools. It teaches the agent a fixed workflow: probe the footage, edit losslessly where possible, then check and verify. It ships 42 tools that do the work with local ffmpeg and ffprobe, covering cutting, joining, silence removal, fitting to a duration or aspect ratio, captions and karaoke captions, HDR to SDR, audio clean-up, loudness and delivery checks, and more. What sets it apart is that it needs no cloud and no API keys, so footage does not have to be uploaded.
Highlights:
- 42 tools that do the work with local ffmpeg and ffprobe
- A fixed workflow: probe, edit losslessly where possible, check, verify
- Every tool can also be called as an MCP tool
- No cloud and no API keys
How to use:
- First, run `npx ffmpeg-skill`.
- Then ask your agent to edit footage, for example to remove silence or recrop to vertical.
- Finally, have it check its output.
Who it's for:
People who want an AI agent to edit video locally without uploading the footage.
Before you start:
- You need ffmpeg and python3 on your PATH.
- Footage is processed locally, with no account or API key.
- The README's demos use synthetic footage that you can rebuild yourself.
- License: MIT.
Based on the project's public documentation. Not tested by AI Inspiration World.