TL;DR : Automatic backup of your camera roll, albums, search, face recognition: Immich does it all like Google Photos does, but on your own server. Your photos stay with you, they're not used to train any model, and you don't pay a subscription to access your own memories. In exchange, you (or we) operate the server and its backups.
Your phone keeps showing « storage almost full ». Google kindly offers a monthly subscription to keep your own photos. Meanwhile those same images feed AI models you don't control. There's another path, and it's matured nicely: Immich.
The idea
Immich is a photo and video management app you install on your own server. The mobile app automatically backs up your camera roll, exactly like Google Photos would, except the destination is a machine you own. The rest of the family or team has their own account, with their own albums and shares.
The difference isn't ideological, it's concrete: no subscription that climbs with volume, no scanning of your images for advertising or training, and the certainty of knowing exactly where your memories live.
What it actually does
Beyond automatic backup, Immich offers a timeline, shared albums, keyword search and face recognition. You get the essentials of the experience the giants' services have trained us to expect, without the privacy trade-off.
How we set it up to be reliable
Technically, Immich runs in containers with a database: we install it on a server in Quebec, behind a reverse proxy with a valid certificate, and configure the mobile app for background backup. Since photos are precious and irreplaceable, we apply the 3-2-1 backup rule from the start: the Immich copy isn't a backup in itself, it's a library that itself needs to be backed up.
Our way of setting up Immich so it stays reliable:
- Dedicated server with enough storage for growth (video gets big fast)
- Automated, off-site backup of the library AND the database
- Encrypted access, per-person accounts, and a restore test before trusting it
Trade-offs to know about
Immich reached its 2.0 stable milestone in the fall of 2025, its very first « stable » release after nearly four years of development: the project adopted semantic versioning, so updates within a given series no longer break your configuration and the mobile app stays compatible with the server. It has since moved on to a 3.0 series (summer 2026). That said, it's still evolving quickly: we keep an eye on the release notes, especially on a major version jump. It's not a service you install and forget for five years. Face recognition and search also need a server with a bit of muscle. And we'll insist on this: Immich centralizes your photos, it doesn't back them up on its own. Without a copy strategy, a failing disk takes everything with it.
The right fit
Immich is ready for anyone who wants to pull their photos back from the cloud without giving up the comfort. For a tech-savvy family, it's a weekend project. For an organization that wants a reliable service without dealing with the admin, it's exactly the kind of thing we host and maintain. The important part: don't confuse « my photos are on my server » with « my photos are safe ». The two are built together.
Ready to pull your memories out of someone else's cloud? Let's set it up for you.
Sources
- Immich : official site and documentation
- Immich on GitHub : code and release notes