Self-hosted Google Photos alternative: own your photo library with Immich

Your photos are the one dataset you can never regenerate. Every birthday, every trip, every face that's changed since — it all lives in one library, and for most people that library sits on Google's servers, counted against a 15GB quota that fills up and then quietly asks for a subscription. Google scans what you upload, and every so often someone wakes up locked out of a decade of memories over an automated flag. Immich is the alternative you actually own: the same automatic phone backup and smart search, running on hardware that answers to you.

What Immich is

Immich is a full self-hosted photo and video platform — not a stripped-down gallery, but a genuine Google Photos replacement built by people who wanted their memories off the cloud without losing the features that made the cloud convenient. It has native iOS and Android apps that back up your camera roll automatically in the background, a fast web timeline, machine-learning search that understands faces and objects, shared albums, and a map of everywhere you've been. Where Google Photos meters you at 15GB and then sells you storage forever, Immich is limited only by the disk you point it at. It handles RAW files, videos, Live Photos and motion photos, reads EXIF metadata, and organizes everything on a timeline that feels immediately familiar. It keeps your originals exactly as you shot them — no quality tier, no re-encoding — so the file you back up is the file you took. It's actively developed, moves fast, and is used daily by tens of thousands of people to hold their entire photographic history.

The features that actually matter

• Automatic background backup — The mobile app uploads new photos and videos the moment you take them, over Wi-Fi or cellular on your terms, with selectable albums and background sync. This is the Google Photos habit, kept.
• Smart search — A CLIP-based model lets you search your library in plain language: "red car," "beach at sunset," "birthday cake." No tags required. It's the same trick Google runs by scanning your library — here it runs locally.
• Face recognition — Immich detects and groups faces, so you can name a person once and pull up every photo of them across years. Merging and hiding faces stays manual and under your control.
• Shared albums and partner sharing — Invite others into an album, or share your whole library with a partner. Public links can carry passwords and expiry dates.
• Accounts for the whole household — One instance hosts everyone: each person gets their own login, library, and timeline, and you decide who can share with whom. It's a real family replacement, not just a personal one.
• Map and metadata — Every geotagged photo lands on a map, and full EXIF — camera, lens, location, timestamp — stays searchable.
• Memories — An "on this day" view surfaces what you shot on the same date in years past, the feature people quietly love most about Google Photos.
• Duplicate detection and stacks — Immich finds duplicates so you can reclaim space, and stacks related shots so a burst of near-identical frames reads as one.

Why self-host it instead of the hosted version

Your library never leaves your hardware. Photos are among the most personal data you own — faces of your kids, where you live, where you were on any given night. Self-hosted Immich means the machine learning runs locally and nobody scans, trains on, or flags your library. There is no terms-of-service that can lock you out of it.

The cost doesn't scale with your life. Google One charges monthly, forever, and the bill grows every time your library does. A self-hosted instance costs the disk once. Twenty thousand photos or two hundred thousand, it's the same free, open-source app.

You control the whole thing. Storage location, retention, who has access, which albums sync — all yours. You can point Immich at existing folders as read-only external libraries, export everything at any time, and back it up however you like. Nothing about your memories is a black box.

Who it's for

We host Immich for three kinds of people. Families who've outgrown the free tier and refuse to rent space for their own baby photos every month. Photographers and hobbyists shooting RAW who want a searchable, geotagged archive without handing full-resolution originals to a cloud that recompresses them. And privacy-minded households who simply don't want a trillion-dollar advertising company holding the faces of everyone they love. If your camera roll is your memory and you'd like to keep it that way, Immich is for you.

The honest setup notes

We'll be straight with you: Immich is not a single container you forget about. It's a small stack — the server, a PostgreSQL database with a vector extension for search, a Redis cache, and a separate machine-learning container that does the face and object recognition. That ML step is genuinely resource-hungry; the first import of a large library will peg a CPU for a while as it processes every photo. Immich also ships frequent updates, and because it's young and fast-moving, its own developers tell you to follow the 3-2-1 rule and never treat it as your only copy of anything. That means real backups of both the database and the originals, on a schedule, tested. Releases occasionally carry breaking changes, too, so an upgrade is something you read the notes for, not something you run blind.

None of that is hard, exactly. It's just ongoing — the database, the updates, the storage headroom, the backups. Or let us run it. We handle the stack, the updates, and the resources; you open the app and see your photos.

Deploy it in one click

Rolling this yourself means a Docker Compose file with four services, a database you have to tune and back up, a machine-learning container to resource correctly, environment secrets to manage, and a reverse proxy with TLS in front of it all. It's doable, and plenty of people enjoy doing it. It's also a weekend and a standing maintenance commitment.

On Caliber Node it's one click. Pick Immich from the catalog, choose your region, and we stand up the whole stack — database, cache, ML, server, TLS — pre-wired and ready. You get a real login on a real domain in a few minutes, with the resources the ML step actually needs already allocated. You can deploy Immich in one click and have your phone backing up tonight.

First run

1. Open your new Immich URL and create the admin account — the first user you register owns the instance.
2. Install the Immich app on your phone from the App Store or Google Play, point it at your server's address, and log in.
3. In the app, open backup settings, choose which albums to sync, and turn on background backup — your camera roll starts uploading.
4. Back on the web, let the machine-learning jobs run through your library, then open Search and try a plain-language query, or open People to start naming faces.
5. Create your first shared album, add a few photos, and send the link to someone — that's the whole loop, working.

The bigger picture

Photos are the closest thing most of us have to a memory we can hold. Handing that to a company that meters it, scans it, and reserves the right to close the door is a strange bargain we only accepted because it was easy. Owning it isn't nostalgia for the shoebox of prints — it's keeping the convenience of the cloud, the automatic backup and the search that actually works, on hardware that answers to you. Immich proves you don't have to choose between features and ownership.

Deploy Immich in a few minutes →

Or browse the full app catalog.

One-click self-hosted apps, managed for you — real logs, SFTP, and your data always exportable. Never a black box.