video-understand — Claude Code skill: what it does, install
This skill extracts video frames and transcribes audio locally with ffmpeg and Whisper. Use it to inspect video content without API keys.
npx skills add calesthio/OpenMontage --skill video-understand -a claude-codeWhat video-understand does
The skill accepts a video file and produces JSON with its duration, resolution, extraction mode, frames, and transcript. It can detect scene changes, extract codec keyframes, or sample frames at even intervals. When scene detection finds no changes, it falls back to interval extraction.
The package includes the Python script skills/video-understand/scripts/understand_video.py. It can cap the number of frames, select a Whisper model, skip transcription, write JSON to a file, or emit JSON without progress messages. Frame extraction uses ffmpeg and ffprobe; audio transcription uses openai-whisper. Plan for ffmpeg and ffprobe before running it. Whisper is only needed for transcript output.
The JSON includes frame paths and timestamps, so an agent can inspect the images with the Read tool. Use -m keyframe for codec keyframes, -m interval for predictable sampling, or leave the scene mode in place for scene-change detection.
What it helps with
- extract key frames from a video
- transcribe video audio locally
- return video metadata and content as JSON
- sample frames at regular intervals
- skip transcription while keeping extracted frames
Example prompts for the agent
- Analyze `video.mp4`, detect scene changes, and transcribe its audio.
- Run `python3 skills/video-understand/scripts/understand_video.py video.mp4 -m keyframe` and inspect the extracted frames.
- Process `video.mp4` without transcription and save the JSON result to `result.json` using `--no-transcribe -o result.json`.
How to install video-understand
1Skills CLI (Vercel)
npx skills add calesthio/OpenMontage --skill video-understand -a claude-code2Manually: put the files in place
~/.claude/skills/video-understand/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/video-understand/
1Skills CLI (Vercel)
npx skills add calesthio/OpenMontage --skill video-understand -a codex2Manually: put the files in place
.agents/skills/video-understand/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add calesthio/OpenMontage --skill video-understand -a cursor2Manually: put the files in place
.cursor/skills/video-understand/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install calesthio/OpenMontage video-understand2Manually: put the files in place
.github/skills/video-understand/SKILL.md1Skills CLI (Vercel)
npx skills add calesthio/OpenMontage --skill video-understand -a gemini-cli2Platform CLI
gemini skills install https://github.com/calesthio/OpenMontage --path .claude/skills/video-understandManually: put the files in place
.devin/skills/video-understand/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add calesthio/OpenMontage --skill video-understand -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
From the same set28
The whole set →Similar
Questions about video-understand
How do I install video-understand in Claude Code?
Run npx skills add calesthio/OpenMontage --skill video-understand -a claude-code, then run npx skills add calesthio/OpenMontage --skill video-understand -a claude-code again. The skill file is at ~/.claude/skills/video-understand/SKILL.md.
What does video-understand do?
It extracts frames with ffmpeg, transcribes audio with Whisper, and returns video details as JSON. Processing is local.
What is required for video transcription?
The script requires ffmpeg and ffprobe. Transcription also needs the openai-whisper package; frame extraction can run with --no-transcribe.