Enables coding agents to watch and analyze videos by extracting scene-aware frames, transcribing speech, and generating shot timelines, all constrained by a token budget to fit LLM context limits.
Enables AI assistants to watch YouTube videos by extracting frames at scene changes and visual references, pairing each frame with the exact words spoken at that timestamp. Provides dense frame-transcript interleaving for any model.
Extracts ffprobe metadata, subtitles, scenes, and timelines from video files without frame-by-frame LLM vision, providing evidence-first reading for AI agents.
Enables AI agents to extract video frames at configurable FPS and compose them into grid sheets for multimodal LLMs, with support for video download, HLS streams, and job management via MCP tools.
Lets any LLM agent actually watch videos: a watch_video tool takes a URL or local file and returns scene-aware keyframes fused with a timestamped transcript, processed 100% locally with per-source caching.