Skip to content

Local Video Editor (FFmpeg)

by Shogo KajiyamaSkillMake video or audio

Lets your agent cut, caption and reframe videos on your own computer with FFmpeg

View on GitHub

The problem it solves

You recorded a 40-minute talk and need a captioned 60-second vertical clip for Reels, but a full video editor is more than you want to learn.

For Short-video creatorsPodcastersSocial media teams

Before
You scrub through a timeline, guess export settings for each app and find out after posting that the captions sit under the buttons.
After
You describe the edit; your agent cuts, captions, reframes and exports the file on your computer, then reports its length and size with a contact sheet.
Stars
★ 4.1k
License
MIT
Cost
Free
Level
Intermediate
Contents
Includes scripts

Examples

From the project's own repository.

Before and after: test footage colour-corrected with a text overlay added by the tools
Image: Shogo Kajiyama (MIT). Original file

Overview

ffmpeg-skill gives your AI agent 42 video and audio tools built on FFmpeg, the free media program. You ask in plain words, for example to keep two parts of an interview and make a 60-second vertical clip; the agent measures the file, edits it, exports it for the platform and checks the result before it reports back. Everything runs on your computer, with no cloud upload and no API key. It suits creators, podcasters and social media teams who edit short clips.

It checks its own output, measuring length, size and loudness and making a contact sheet to look at, instead of just saying it is done.

Try these first

Start with the first one: it is a small task you can check easily.

  1. Take [clip.mp4], keep 10 seconds starting at [0:30] and export it as 9:16. Tell me the final length and size.First try · writes a new file

  2. Remove the silences from [podcast.wav] and normalise it to -14 LUFS.writes a new file

  3. Add these captions to [talk.mp4] for TikTok: [paste timed captions]writes a new file

  4. Check whether [final.mp4] meets YouTube's upload specs. Don't change it.read-only

Example prompts written by Agent Flows Hub.

What you get

From the creator's documentation.

OutputWhat it isReads
Cut and joinTrim sections, join clips and remove silences, without re-encoding where possibleYour files
ReframeChange the shape to 9:16 or 1:1 by cropping or paddingYour files
Captions and titlesBurned-in captions, animated or karaoke style, plus logos and lower-thirdsYour files
AudioLoudness levelling, background music with ducking and external-mic syncYour files
Platform exportPresets and spec checks for YouTube, Reels, TikTok and XYour computer
Result checkMeasures the output and makes a contact sheet for the agent to look atYour computer

Good to know

Cut a long recording into a short vertical clip with captions and the right loudness for a platform. The skill measures the file first, runs fixed editing tools and checks the exported file against the platform's specs.

Limits

  • FFmpeg is installed separately and needs the filters the tools use; on macOS the plain Homebrew ffmpeg lacks subtitles and text drawing. Captions from speech need a separate local speech engine such as faster-whisper, which downloads a model on first use.
  • The tools carry out edits but cannot judge them. Whether a face is cut off or text covers a subject is for you or your agent to check, using the contact sheet the skill makes.

Notes by (original project name: ffmpeg-skill), written from the creator's documentation, not from running it ourselves. ffmpeg-skill README: quick start, install and requirements · ffmpeg-skill: SKILL.md

Stuck setting it up?

Tell us what happened. We take a look first; if it needs paid work, you get an estimate before anything starts.

Get help

Other skills

See all 140 skills · A skill is installed but not loading

GitHub