Gives Claude five tools to break down video content without leaving the chat. Point it at a YouTube URL or local file and video_analyze runs the full pipeline: scene detection pulls keyframes, Whisper transcribes the audio with timestamps, Claude Vision describes each frame, and you get back an eight-field stylistic fingerprint classifying everything from rendering style to tonal positioning. You can also call the steps individually if you just need transcription or frame extraction. Requires FFmpeg locally and an Anthropic API key in the environment. Outputs storyboards as Markdown or DOCX, useful when you're analyzing reference videos, auditing content at scale, or feeding visual context into longer creative briefs.
claude mcp add --transport stdio brightwayai-video-analyzer -- uvx video-analyzer-mcp