PhotoPrism

PhotoPrism is an open-source, privacy-first photo and video management app that runs on your own hardware (Docker, NAS, Raspberry Pi, FreeBSD, or a small cloud VM). It uses on-device machine learning to do the things Google Photos and iCloud do in the cloud: face recognition for tagging family and friends, automatic content classification (beach, dog, document, sunset, etc.), location clustering on an interactive world map, and full-text search across that derived metadata. It handles RAW, Live Photos, sidecar files, and duplicate detection out of the box.

It is aimed at photographers, families, and homelab users who want a real alternative to Google Photos without uploading their library to a third party. The core software is free and AGPL-licensed; the project funds itself through a membership program (Silver, Gold, Platinum tiers) that unlocks extras and supports development, plus a separate PhotoPrism Pro for teams. Compared to Immich, the other big contender in this niche, PhotoPrism leans more mature, slightly heavier, and more focused on classification and discovery than on mobile-first backup.

There is a documented REST API, an official mobile app for members, and integrations with WebDAV clients so you can sync from phones and desktops. The main caveats are typical of self-hosting: you are responsible for backups, GPU acceleration is optional but recommended for large libraries, and the indexing pass on a fresh import can be slow on a Pi-class device.