Immich vs PhotoPrism and Nextcloud: setup and comparison

The original video is in German. The presenter compares Immich, PhotoPrism and Nextcloud as alternatives to Google Photos and Apple Photos. He reports moving a library of roughly 700 GB to his own hardware, with local AI handling photo search and facial recognition. His search examples return useful matches but also false positives, so the comparison reflects his library and tests.

Immich gets most of the setup coverage: Docker installation, browser access, phone backup, exports through Google Takeout and bulk imports with its CLI. The tutorial also covers duplicate detection, family accounts and shared links. On a Raspberry Pi 4B with 8 GB RAM and a USB SSD, the presenter found image processing slow. He suggests disabling machine learning or running that service on a separate PC or Mac. He notes that Immich lacks end-to-end encryption and recommends HTTPS for remote access.

PhotoPrism suits existing folder archives in this comparison. The presenter demonstrates metadata filters, RAW support, related-file stacking and a read-only mode that leaves originals untouched. Its Docker setup requires attention to storage paths and the administrator password. For existing Nextcloud users, he recommends Memories for its thumbnail timeline and Recognize for facial recognition and object tagging.

The closing advice concerns backups: a self-hosted photo library still needs a separate backup, because RAID alone does not protect against deletion or other data loss.