Immich Self Hosted Docker: 3 Reasons It Beats Google Photos

So I’ve been paying for Google Photos storage for years, and it finally hit me: I’m essentially renting space for my own memories on someone else’s server. When I stumbled onto immich self hosted docker a few months back, it changed everything. I set it up on a Raspberry Pi 5 in an afternoon, and I haven’t looked back since.

The wild part? It’s not just a basic file dump. Immich has face recognition, AI-powered search, automatic mobile backup, and a timeline view that honestly looks better than Google Photos. And it all runs on hardware you own, in your house, forever — no subscription, no storage caps, no terms of service that can change overnight.

Transparency Note: This post contains affiliate links. If you buy something through these links, I may earn a small commission at no extra cost to you. Every product mentioned is researched based on specs, expert reviews, and real user feedback.
immich self hosted docker running on a Raspberry Pi home server

Aspect Google Photos Immich Self-Hosted
Storage limit Tiered plans, capped storage Whatever drives you attach
Ongoing cost Monthly/annual subscription One-time hardware cost, then free
Data location Google’s servers Your own hardware
AI features (face recognition, smart search) Yes, cloud-based Yes, runs locally (CLIP-based natural language search)

What Is Immich Self Hosted Docker and Why Does It Matter?

Immich self hosted Docker deployments have become the default Google Photos exit route — here’s why the switch finally makes sense.

Immich is a self-hosted photo and video management platform that runs as a Docker Compose stack. It started as a hobby project a few years ago and has exploded into one of the most actively developed self-hosted apps out there. As of 2026, it’s rock-solid, has an official app for iOS and Android, and supports everything from facial clustering to CLIP-based natural language search (yes, you can literally search “birthday cake” and it’ll find the right photos).

The Docker Compose setup spins up four containers: the main server, a machine learning service for AI features, a Valkey cache, and a Postgres database with vector extensions. Sounds complicated, but Immich ships an official docker-compose.yml that handles all of it. You copy two files, fill in one environment variable (your upload path), and run docker compose up -d. That’s it.

Here’s why it beats paying Google month after month:

  • No storage limits — capacity is whatever drives you attach
  • No subscription — one-time hardware cost, then it’s free forever
  • Your data stays home — GDPR-proof, Cloud Act-proof, terms-change-proof
  • AI features built in — face recognition, object detection, smart search — all local

Reason 1 — The Hardware Is Cheaper Than You Think

You don’t need a rack server or a dedicated NAS appliance to run Immich. A Raspberry Pi 5 (8GB) handles it comfortably for a household photo library. Immich recommends a minimum of 6GB RAM — the Pi 5’s 8GB model hits that with room to spare, and its quad-core Cortex-A76 processor handles the machine learning inference faster than the Pi 4 ever could.

I run mine 24/7 and it idles at about 8W of power draw. Compare that to a desktop PC running as a server (60–150W) and the Pi pays for itself in electricity savings within a year. The key is pairing it with an SSD for the OS and database — SD cards die under constant write load. The Samsung T7 1TB USB SSD is what I use: fast, compact, and plugs right into the Pi’s USB 3.0 port.

The Hardware Is Cheaper Than You Think

Raspberry Pi 5

Enough machine for a full Immich stack, at a power draw you stop noticing.

  • Quad-core A76 handles ML inference well
  • 8GB RAM leaves headroom for transcoding
  • Tiny and silent
  • Trade-off: Not for 4K transcoding or libraries over 10TB — step up to an N100 mini PC at that scale.

Check price on Amazon →

Current price and availability shown on Amazon.


Reason 2 — Storage Is Dirt Cheap When You Own It

This is where the immich self hosted docker setup really shines. You mount whatever storage you want — a USB hard drive, a NAS share, a folder on your SSD — and Immich just uses it. No tiered plans, no “you’ve hit your limit” emails.

For storing actual photos, I recommend a purpose-built NAS drive rather than a desktop or portable drive. The WD Red Plus 4TB NAS Hard Drive is built for always-on workloads — it has a 256MB cache, CMR recording (important for data integrity in 24/7 use), and a 3-year warranty. At 4TB you can store roughly 800,000 high-res JPEG photos or several years of 4K video. Most people never fill a single drive.

In your .env file, you set one line: UPLOAD_LOCATION=/mnt/photos. Mount the drive there, and every photo uploaded from the mobile app or the web interface lands directly on that drive. It’s shockingly straightforward.

Storage Is Dirt Cheap When You Own It

WD Red Plus

The drive the photos actually live on, rated for the 24/7 duty this really is.

  • CMR recording for data integrity
  • 256MB cache for burst writes
  • NAS-rated for 24/7 operation
  • Trade-off: Needs a USB enclosure on a Raspberry Pi, or a NAS box if you want RAID.

Check price on Amazon →

Current price and availability shown on Amazon.


Reason 3 — The Setup Is Genuinely Quick

I was bracing for a weekend-long project. It took about 90 minutes, and half of that was waiting for Docker to pull images. Here’s the gist of how an immich self hosted docker deployment works:

First, install Docker and Docker Compose on your Pi (one command on Raspberry Pi OS). Then grab Immich’s official docker-compose.yml and .env from immich.app — they maintain these files and update them with every release. Set UPLOAD_LOCATION to point at your mounted drive, set DB_PASSWORD to something secure, and run:

docker compose up -d

Docker pulls four images and starts the stack. When it’s ready (about 2–5 minutes), open http://your-pi-ip:2283 in a browser. You’ll see the Immich admin registration page. Create your account, download the iOS or Android app, point it at your server’s IP, and flip on automatic backup. Done.

The mobile app detects new photos in the background and uploads them over Wi-Fi. Within a few hours, your library syncs. Immich then runs its machine learning pass in the background — face clustering, CLIP embeddings, object tagging — and after a day or so, your search is fully operational. Type “Christmas 2023” or “dog at the beach” and it’ll find the right photos.

For the SSD (OS and database), the Samsung T7 1TB plugs into the Pi’s USB 3.0 port and boots reliably without any fussing. Pair it with a USB hard drive enclosure for the WD Red and you’ve got a complete, quiet, always-on photo server for a modest hardware outlay.

The Setup Is Genuinely Quick

Samsung T7 Portable SSD

The OS and database drive that keeps the Pi off an SD card.

  • Plugs straight into the Pi’s USB 3.0 port
  • Fast and compact
  • Boots reliably with no fussing

Check price on Amazon →

Current price and availability shown on Amazon.

One Caveat Before You Commit: You Are the Backup Plan Now

Here’s the part most immich self hosted docker guides skip: when you leave Google Photos, you also leave Google’s disaster recovery team. If that WD Red dies with the only copy of your kid’s first birthday on it, there is no support ticket to file. So build the backup habit into day one. Immich already dumps its database nightly into your upload folder, which means everything that matters — photos, albums, face data — lives in one directory tree you can copy like any other files.

My setup is the classic 3-2-1: the live copy on the Pi, an automated nightly rsync to a second USB drive, and a weekly encrypted push to cloud object storage for the offsite copy. If you’re already running Proxmox elsewhere in your homelab, my Proxmox Backup Server setup guide covers the same idea with deduplication built in. Ten minutes of cron configuration buys you the one thing Google was actually charging you for: the certainty that the photos survive the hardware.


The Takeaway

Immich self hosted Docker photo hosting means your memories live on your hardware, indexed by AI that reports to no one.

If you’re already running any kind of home server — or even if you’re not — immich self hosted docker is the most satisfying upgrade I’ve made to my personal data setup in years. You get Google Photos-level features, zero ongoing costs, and the satisfaction of knowing your family photos aren’t locked inside some cloud company’s data center. The hardware cost is a one-time hit. The freedom is permanent.

The Raspberry Pi 5, a Samsung T7 for the OS, and a WD Red Plus for storage will run you a modest one-time total. That’s less than two years of a Google One 2TB plan — and this doesn’t expire.

Got Immich running on your setup? Drop your specs in the comments — I’d love to see what people are running this on. If you’re on a mini PC or a full NAS, I especially want to hear how the ML performance compares.

Frequently Asked Questions

How much hardware do I need to self-host Immich?

Not much — a Raspberry Pi 5 (8GB) handles a household photo library comfortably. Immich recommends a minimum of 6GB RAM, and the Pi 5’s quad-core Cortex-A76 processor is fast enough for the machine learning features.

How long does it take to set up Immich with Docker?

About 90 minutes, and roughly half of that is just waiting for Docker to pull the four container images. The official docker-compose.yml and .env files from immich.app handle most of the configuration for you.

Is my photo data safe if I switch from Google Photos to Immich?

It’s safe as long as you build backups in yourself — once you leave Google Photos, you also leave Google’s disaster recovery team. A simple 3-2-1 setup (a live copy, an automated nightly sync to a second drive, and an offsite copy) covers this in about ten minutes of cron configuration.

How much does a self-hosted Immich setup cost?

A Raspberry Pi 5, an SSD for the OS and database, and a NAS-rated hard drive for photo storage — a modest one-time total. That’s less than two years of a Google One 2TB plan, and the hardware doesn’t expire.

Does Immich have AI-powered photo search like Google Photos?

Yes — face recognition, object detection, and CLIP-based natural language search all run locally. You can type something like “dog at the beach” and it finds the matching photos, the same way Google Photos does, just without the data leaving your house.

Scroll to Top