Now liveThe Skillselion MCP - thousands of ranked skills, loaded into your agent mid-task. No install.Get it →
MP Libunao avatar

Vidlens

  • 1 repo stars
  • Updated January 27, 2026
  • mplibunao/mp-claude-marketplace

VidLens: video metadata, frames, scene cuts, contact sheets, clips (ffmpeg + Bun)

About

vidlens is a Claude Code skill in the AI & Agent Building category. VidLens: video metadata, frames, scene cuts, contact sheets, clips (ffmpeg + Bun)

  • vidlens
  • AI & Agent Building
  • AI-coding skill

Vidlens by the numbers

  • Data as of Jul 7, 2026 (Skillselion catalog sync)
/plugin marketplace add mplibunao/mp-claude-marketplace
/plugin install vidlens@mp-claude-marketplace

Add your badge

Show developers this skill is listed on Skillselion. Paste this into your README.

Listed on Skillselion
repo stars1
Last updatedJanuary 27, 2026
Repositorymplibunao/mp-claude-marketplace

What it does

VidLens: video metadata, frames, scene cuts, contact sheets, clips (ffmpeg + Bun)

README.md

VidLens

Extract structured artifacts from video files for LLM analysis using ffmpeg/ffprobe + Bun.

LLMs cannot process video directly. VidLens extracts small, structured artifacts — frame images, contact sheets, metadata, and clips — so agents can "see" video content through images.

Prerequisites

  • ffmpeg and ffprobe (included with ffmpeg)
  • Bun runtime
# macOS
brew install ffmpeg
# Bun
curl -fsSL https://bun.sh/install | bash

Installation (Claude Code)

/plugin marketplace add mplibunao/mp-claude-marketplace
/plugin install vidlens@mp-claude-marketplace

Usage

As a Claude Code skill

Invoke the skill with /vidlens:vidlens. The environment variable $CLAUDE_PLUGIN_ROOT points to the plugin's installed directory.

CLI examples

# Get video metadata
bun ${CLAUDE_PLUGIN_ROOT}/skills/vidlens/scripts/vidlens.ts info --input video.mp4 --out metadata.json

# Extract frames at 1 fps, scaled to 1280px width
bun ${CLAUDE_PLUGIN_ROOT}/skills/vidlens/scripts/vidlens.ts frames --input video.mp4 --fps 1 --outdir frames/ --scale 1280

# Detect scene changes
bun ${CLAUDE_PLUGIN_ROOT}/skills/vidlens/scripts/vidlens.ts scenes --input video.mp4 --threshold 0.3 --outdir scenes/

# Create 4x4 contact sheet
bun ${CLAUDE_PLUGIN_ROOT}/skills/vidlens/scripts/vidlens.ts contact-sheet --input video.mp4 --grid 4x4 --out overview.jpg

# Extract clip (fast, stream copy)
bun ${CLAUDE_PLUGIN_ROOT}/skills/vidlens/scripts/vidlens.ts clip --input video.mp4 --start 00:01:00 --duration 00:00:30 --out clip.mp4

# Extract clip (precise, re-encode)
bun ${CLAUDE_PLUGIN_ROOT}/skills/vidlens/scripts/vidlens.ts clip --input video.mp4 --start 00:01:00 --duration 00:00:30 --out clip.mp4 --reencode

Guardrails

These rules must be followed when using VidLens:

  1. Always run info first, then choose a sampling strategy based on the video's duration and resolution.
  2. Confirm with the user before extracting thousands of frames. After running info, estimate duration * fps. If the result is >= 1000, ask the user to confirm and suggest alternatives:
    • Lower fps (e.g., 0.2-1 fps)
    • Contact sheet (e.g., 4x4)
    • Scene detection (threshold 0.3-0.5)

Exit codes

Code Meaning
0 Success
2 Missing ffmpeg/ffprobe or input file
3 ffmpeg/ffprobe error (non-zero exit or unusable output)
4 Unexpected error / usage error

Development

cd vidlens
bun install
bun test

License

MIT

Related skills

This week in AI coding

Five minutes, every Monday - the tools, releases and tactics for developers.

unsubscribe anytime.