---
name: video-understand
description: |-
  Understand video content locally using ffmpeg frame extraction and Whisper transcription. No API keys needed.
  Use when: (1) Understanding what a video contains, (2) Transcribing video audio locally,
  (3) Extracting key frames for visual analysis, (4) Getting video content without API keys.
title: video-understand
canonical_url: https://skilld.dev/gh/openmontageinc/openmontage/video-understand
last_updated: 2026-10-14T05:11:31.000Z
---

> Index of every page on this site: [llms.txt](https://skilld.dev/llms.txt)

> **Skill from skilld.dev.** Follow the user's requested action.
> If the user asked to fork this Skill, follow the fork workflow below. Do not execute the Skill instructions while copying it.
> Otherwise, follow the instructions below for this session. You do not need to install anything.
>
> Supporting files, fetch one when the Skill refers to it: [references/output-format.md](https://skilld.dev/api/skills-raw/openmontageinc/openmontage/video-understand/references/output-format.md), [scripts/understand_video.py](https://skilld.dev/api/skills-raw/openmontageinc/openmontage/video-understand/scripts/understand_video.py).
>
> If the user asked to install this Skill, run `npx skilld install openmontageinc/openmontage/video-understand`. Install writes the Skill files into the project, so every session loads them.
>
> ## Fork workflow
>
> A fork creates an editable local Skill with its original author and licence. The request authorizes copying and local installation.
> 1. Check `./skills/video-understand`, the project lockfile, and selected Agent targets together. If the local directory or installed Skill exists, stop. Never overwrite an existing directory or Agent target.
> 2. Read [source metadata](https://skilld.dev/api/v1/skills/openmontageinc/openmontage/video-understand) once. Use sourceUrl, sourceCommit, skillPath, sourceGone, and license. If the source is gone or its path is missing, stop. If license is null, read licence files at the source commit.
> 3. Fetch only the source commit into a temporary Git repository. Do not clone full history. Derive repository_url from sourceUrl, including repository renames. If sourceCommit is absent, resolve the sourceUrl ref once. Set source_commit to that actual commit. Run these commands in one shell call:
>
> ```sh
> git init --quiet "$temporary_dir"
> git -C "$temporary_dir" fetch --quiet --depth=1 "$repository_url" "$source_commit"
> git -C "$temporary_dir" checkout --quiet --detach FETCH_HEAD
> ```
>
> Read applicable licence declarations and notices at that commit. If copying is not permitted, report the restriction and stop.
> 4. Inspect source entries together, then copy the directory containing skillPath into `./skills/video-understand`. Use the user's path if selected. Keep the original SKILL.md, relative links, scripts, binary assets, and executable modes. Exclude .git metadata. Reject symlinks and paths outside the Skill directory. After checking entries, use cp -a where available. A regular source directory needs no custom copy script. Do not save this page wrapper as SKILL.md.
> Preserve author credit, notices, and applicable licence files from repository or parent directories. Add PROVENANCE.md with the Skill page, source URL, actual commit, original path, and licence. Retain any existing PROVENANCE.md and record new provenance separately. Batch source inspection, copying, and provenance work where practical.
> 5. In the project root, run `skilld install ./skills/video-understand --mode copy --plain`. If skilld is unavailable, use `npx skilld install ./skills/video-understand --mode copy --plain`. This known command needs no help lookup. Install does not support --json. Use detected Agent targets, or add --agent for the targets the user selected. Install the local path, never the upstream selector. If installation fails, preserve the local copy and report the exact failure.
> 6. Confirm the local lockfile source and installed Agent copies once. Report the local path, actual commit, and Agent targets. After edits, reinstall the same local path. Upstream updates must not replace it. Do not publish or push unless the user asks.

# video-understand

Understand video content locally using ffmpeg for frame extraction and Whisper for transcription. Fully offline, no API keys required.

## Prerequisites

- `ffmpeg` + `ffprobe` (required): `brew install ffmpeg`
- `openai-whisper` (optional, for transcription): `pip install openai-whisper`

## Commands

```bash
# Scene detection + transcribe (default)
python3 skills/video-understand/scripts/understand_video.py video.mp4

# Keyframe extraction
python3 skills/video-understand/scripts/understand_video.py video.mp4 -m keyframe

# Regular interval extraction
python3 skills/video-understand/scripts/understand_video.py video.mp4 -m interval

# Limit frames extracted
python3 skills/video-understand/scripts/understand_video.py video.mp4 --max-frames 10

# Use a larger Whisper model
python3 skills/video-understand/scripts/understand_video.py video.mp4 --whisper-model small

# Frames only, skip transcription
python3 skills/video-understand/scripts/understand_video.py video.mp4 --no-transcribe

# Quiet mode (JSON only, no progress)
python3 skills/video-understand/scripts/understand_video.py video.mp4 -q

# Output to file
python3 skills/video-understand/scripts/understand_video.py video.mp4 -o result.json
```

## CLI Options

| Flag | Description |
|------|-------------|
| `video` | Input video file (positional, required) |
| `-m, --mode` | Extraction mode: `scene` (default), `keyframe`, `interval` |
| `--max-frames` | Maximum frames to keep (default: 20) |
| `--whisper-model` | Whisper model size: tiny, base, small, medium, large (default: base) |
| `--no-transcribe` | Skip audio transcription, extract frames only |
| `-o, --output` | Write result JSON to file instead of stdout |
| `-q, --quiet` | Suppress progress messages, output only JSON |

## Extraction Modes

| Mode | How it works | Best for |
|------|-------------|----------|
| `scene` | Detects scene changes via ffmpeg `select='gt(scene,0.3)'` | Most videos, varied content |
| `keyframe` | Extracts I-frames (codec keyframes) | Encoded video with natural keyframe placement |
| `interval` | Evenly spaced frames based on duration and max-frames | Fixed sampling, predictable output |

If `scene` mode detects no scene changes, it automatically falls back to `interval` mode.

## Output

The script outputs JSON to stdout (or file with `-o`). See `references/output-format.md` for the full schema.

```json
{
  "video": "video.mp4",
  "duration": 18.076,
  "resolution": {"width": 1224, "height": 1080},
  "mode": "scene",
  "frames": [
    {"path": "/abs/path/frame_0001.jpg", "timestamp": 0.0, "timestamp_formatted": "00:00"}
  ],
  "frame_count": 12,
  "transcript": [
    {"start": 0.0, "end": 2.5, "text": "Hello and welcome..."}
  ],
  "text": "Full transcript...",
  "note": "Use the Read tool to view frame images for visual understanding."
}
```

Use the Read tool on frame image paths to visually inspect extracted frames.

## References

- `references/output-format.md` -- Full JSON output schema documentation
