PhotoPrism

PhotoPrism is a self-hosted application for browsing, organizing, and sharing photos and videos. It uses machine learning to automate the categorization of media libraries, allowing users to manage their collections on their own hardware or in a private cloud environment. The software focuses on privacy by ensuring that data remains under the control of the user rather than being shared with third-party cloud providers.

The software is deployed as a server and accessed via a web browser as a Progressive Web App (PWA), which allows it to be installed on home screens for a native experience on phones and tablets. It is compatible with Mac, Linux, and Windows, and is primarily installed using Docker. The application is available as a multi-arch image, ensuring that it runs on 64-bit AMD, Intel, and ARM processors, including hardware like Raspberry Pi and Apple Silicon.

Key features

• Automatic content labeling and face recognition using machine learning
• Search filters based on labels, location, resolution, color, chroma, and quality
• Support for RAW images and various video formats
• Metadata extraction and merging from Exif, XMP, and Google Photos
• World maps and a privacy-preserving geocoding service for location browsing
• WebDAV support for direct file access via Windows Explorer or Apple Finder
• Integration with PhotoSync for background backups from iOS and Android
• Support for Live Photos with hover-to-play functionality

Built with Golang and TensorFlow, the application implements a privacy-first architecture that avoids external data synchronization. It allows users to maintain their own index of media while providing tools to rediscover old photos through combined search filters. The system is designed for photographers and individuals who want to maintain full ownership of their media metadata and files without relying on proprietary cloud services.

PhotoPrism serves as a private media server for users seeking automated organization and high-resolution mapping of large photo and video archives.