Convert Videos into Searchable, AI-Readable Intelligence

The Video Analyzer skill bridges the gap between raw video footage and your AI workflows. Many LLMs struggle with direct video uploads or offer limited context windows for long files. This skill solves that by locally processing video files into structured Markdown and JSON reports that capture every essential detail.

What it does

This developer-focused tool performs deep multi-modal analysis locally on your machine. It follows a rigorous workflow to decompose video content:

• Automated Pre-processing: Extracts frames and audio using ffmpeg and opencv.
• Speech-to-Text: Generates high-accuracy transcripts using OpenAI's Whisper.
• Computer Vision: Detects distinct characters, identifies visual changes, and generates a visual timeline of events.
• Structured Output: Compiles metadata, transcriptions, and visual descriptions into a final schema-ready JSON and a human-readable Markdown report.

Why use this skill

Prompting an AI to "watch a video" often leads to hallucinations or missing segments. This skill uses deterministic local processing to ensure nothing is missed. It is entirely private—processing happens via local subprocesses with no outbound network calls. The output is perfectly formatted for RAG (Retrieval-Augmented Generation) or for use as context in long-form coding and analysis tasks.

Supported tools

Built for efficiency, it leverages ffmpeg for media handling, Whisper for audio, and OpenCV for frame analysis. It supports all major formats including .mp4, .mov, .avi, .mkv, and .webm.