TL;DR
A tool that gives Claude the ability to 'watch' videos by providing extracted frames and transcribed captions/audio when given a video URL or local file.
Key features
Multi-source support: Handles any URL supported by yt-dlp (YouTube, Loom, TikTok, etc.) and local video files.
Intelligent frame extraction: Extracts key frames efficiently using three modes: efficient (keyframes only), balanced, and token-burner.
Caption/Audio transcription: Checks for captions first, then falls back to Whisper API for audio-to-text conversion.
One-click installation: Easily installed via the Claude Code plugin marketplace or npm.
When to use it
When you need to structurally analyze video content from competitors or popular YouTubers.
When visually diagnosing a bug reported via screen recording.
When you want to quickly summarize the key points of a long lecture or update video.
When organizing a video playlist into searchable notes.