Windrecorder is a personal memory search engine that records everything on your screen locally to let you rewind and query what you have seen.
The tool addresses the problem of retrieving past screen activity by continuously capturing video from your display at low resource cost, then indexing only changed scenes to a local database. It lets you search through recordings using optical character recognition of on-screen text or image descriptions, and provides activity statistics including word clouds, timelines, and scatter plots. All processing happens locally without requiring internet connectivity or uploading data to external services.
Windrecorder suits users who want a self-hosted alternative to commercial memory search applications and need to maintain complete privacy over their screen recordings. It works on Windows and supports multiple OCR engines including Rapid OCR, WeChat OCR, and Tesseract, allowing you to choose based on language support and accuracy requirements. The tool includes a web interface for reviewing recordings and conducting searches, supports multiple languages, and can automatically maintain the database by cleaning and compressing video when the computer is idle. Custom skip conditions let you exclude specific windows, processes, or inactive periods from recording.
Development activity shows consistent engagement with bug fixes and feature improvements across multiple areas. The project maintains responsiveness to user-reported issues and incorporates feedback into updates. Work spans the full stack from video capture and OCR integration to the web interface and database management. The maintainers actively document setup procedures and troubleshooting steps, including detailed guidance for updating and recovering from conflicts.
