Finding a private self hosted Google Photos alternative that doesn’t profit from my photos

Over the last few weeks I have been trying to find a comfortable middle ground between the convenience of mainstream photo apps and the control you get from self hosted tools. Big players like Apple Photos or Google Photos make it incredibly easy to back up and search everything with decent cross device syncing, but that comfort comes with the cost of letting their machine learning models scan your personal memories and faces of your loved ones, which Google even acknowledged in its own documentation. After trying Ente’s TheySeeYourPhotos and seeing how Google’s API could infer locations, political beliefs, salaries or even bits of someone’s life story from a single upload, it became harder to ignore what I was giving away for free.

So I gave myself a small project and tried to build a private photo management setup to see how close it could get to a Google Photos style experience. I wanted my photos away from third party AI training, no ad driven services, tools that were either self hosted or properly encrypted, and nothing that turned me into a part time sysadmin. With that in mind I started exploring the options that kept coming up whenever people talked about private photo libraries.

PhotoPrism and LibrePhotos as your own local AI engines

The first attempt at a full featured gallery that caught my eye was PhotoPrism. It uses several AI models running entirely locally for automatic tagging, meaning my private data is not used to train external models while still giving me features I enjoy like face recognition, tags and maps. The problem? Setting it up means dealing with Docker and a database, and updates sometimes need more care than I would like, although the result is impressive. LibrePhotos felt similar, with a more community driven spirit and fewer sharp edges, plus a timeline view and multiuser support, but still too rough to feel like something I would open every day.

Immich and the “this feels familiar” moment

Immich changed things. It looks and behaves uncannily like Google Photos, just without Google. Backups happen quietly, the timeline feels familiar and features like face clustering and semantic search work without leaving your server. They even recently made version 2.0 available for purchase on a physical DVD! how cool is that? The project moves fast, sometimes faster than I can keep up with, but it was the first self hosted gallery that made me forget I was self hosting at all. Just beware that it is still fairly new and only recently left beta, so you might run into occasional stability issues or breaking changes when upgrading.

Best of both worlds, where Ente fits in

Immich solved most things, but I still wanted an encrypted off site safety net and Ente Photos filled that role. Everything is end-to-end encrypted before it touches their servers, and I can still self host the backend if I want. Like the others, it is open source but benefits from independent security audits that reviewed Photos, Ente Auth and Locker, a nice new little app for storing IDs and sensitive files (not photos) included in their subscription. Speaking of subs, Ente gives you 10 GB of free space and offers plans starting at $2.50 a month, with public links and family sharing.

So where did I land?

I have to admit that after all the testing, I keep circling back to Ente Photos. Immich is fantastic and easily the closest thing to a self hosted Google Photos, but Ente feels more plug and play, predictable and far less demanding on my time. I can still self host it if I want full control, yet its hosted version already gives something closer to the privacy I was looking for. If you go down this path, my advice is simple, start from your real tolerance for maintenance and your actual privacy needs. For many people an Ente account will already be a huge step forward, while for others running Immich or PhotoPrism on a home server will be the part that feels empowering.

Our take: I ran everything on an old laptop I had sitting around, using Docker with a regular folder mounted as my library. Most of these tools include a docker compose file, so it was just pulling it, pointing the paths to my storage and letting docker compose up -d do the work. A Raspberry Pi can handle small libraries, but a bit more CPU helps with indexing and face clustering. This post is not meant to be a setup guide but rather to share the experience of testing these tools, so for a more technical step by step reference you can check the Immich and PhotoPrism documentation.