Video Edit CLI

Turn raw recordings into publish-ready videos with your AI agent. Ask in plain language to edit a podcast, make YouTube Shorts with burned-in subtitles, remove silences and false starts, sync multi-camera shoots, or restore and master audio. The tool is open source, runs locally, and has no watermarks or per-clip credits.

We use this tool at computerlove.tech to edit our video podcasts. We made it open source so your agent can edit yours too.

It works with any coding agent that supports skills: Claude Code · Codex · Cursor · GitHub Copilot · Pi · and others.

Get started

Install the skill into your project with skills.sh:

npx skills add computerlovetech/video-edit-cli --skill video-edit-cli

Then ask your agent to edit a recording:

Make clips for YouTube Shorts from this video, with burned-in subtitles: https://youtu.be/r1Kh5WssSPg

The skill guides the agent through the workflow and installs the CLI when needed. You need only ffmpeg and ffprobe.

More prompts that work well: example prompts. Have a good one? Add it.

What your agent can do with it

• Edit a podcast or talk into a publishable episode — cut false starts, retakes, and dead air from the transcript, keeping the conversation intact.
• Turn a long video into vertical Shorts/Reels — pick moments from the transcript, reframe to 9:16, and burn in styled subtitles.
• Transcribe and subtitle — word-level Whisper transcripts, SRT/VTT export, muxed or burned-in subtitles.
• Restore and master audio — extract, denoise, A/B compare, and loudness-normalize for podcast or YouTube delivery.
• Sync multi-camera and separate-audio shoots — estimate and apply offsets, then cut between angles in one plan.
• Inspect unknown media — check streams, durations, frames, filmstrips, waveforms, and low-resolution previews before editing.

How it works

The CLI provides small, fixed commands. The agent gathers evidence from transcripts, frames, and waveforms. It then writes an edit plan: an ordered keep-list that gives a reason for every cut. The CLI validates and renders the plan:

{
  "plan_id": "rough-cut-1",
  "timeline": [
    {"source": "src-1", "in": 0.5, "out": 42.1, "reason": "keep intro"},
    {"source": "src-1", "in": 55.0, "out": 120.4, "reason": "false start removed at 42.1-55.0"}
  ]
}

Every command prints one JSON result on stdout. Commands never change source media. Each derived file gets a *.provenance.json sidecar that records how the tool made it, so you can trace and reproduce every edit.

video-edit-cli workspace init --root /tmp/ep1 --source recording.mp4
video-edit-cli probe --input recording.mp4
video-edit-cli filmstrip create --input recording.mp4 \
  --start 60 --end 90 --output /tmp/ep1/analysis/strip.png --workspace /tmp/ep1

Install the CLI directly with uv tool install video-edit-cli. For Apple Silicon transcription, run uv tool install 'video-edit-cli[mlx]'. Use --help on any subcommand to list its options. Check dependencies with video-edit-cli doctor.

Runs on macOS and Linux. Windows is untested; use WSL.

Docs

Full documentation: https://computerlovetech.github.io/video-edit-cli/