Claude Code Video Vision
A Claude Code plugin that extracts video frames with ffmpeg and transcribes audio through Gemini, local Whisper, or OpenAI, giving Claude frames as images and timestamped transcripts to answer questions about a video.
Claude Code Video Vision gives Claude the ability to watch and understand videos. It extracts frames through ffmpeg and processes audio through a choice of backends (the Gemini API, local Whisper, or the OpenAI API), then hands Claude the frames as images and the audio as a timestamped transcript; the plugin is described as a perception layer rather than an interpretation layer.
A YouTube URL can be passed directly, and the bundled MCP server downloads it with yt-dlp while preserving source metadata and captions. Claude adjusts frame rate, time range, and resolution based on the question asked, for example sampling at full frame rate for a single second or at low frame rate across a full lecture.
It installs as a Claude Code plugin, with the MCP server auto-installing through npx on first use, and includes an interactive setup wizard for backend and Whisper configuration.