Immich vs Ente

We earn commissions when you shop through the links below. Learn more

Comparison · Immich

Versions/date: Immich v3.3.5 (GitHub release). Ente Photos apps photos-v1.3.66 (GitHub release); Ente's server has no numbered release, so the self-hosting notes follow the Ente help docs as of commit b356bf8 (2026-10-29). Last checked: 2026-10-29.

Both are open-source Google Photos alternatives with mobile apps, face recognition and natural-language search. The big difference is where the work and the keys live. Immich does its indexing on your server and keeps your photos as normal files you can see on disk. Ente encrypts everything on your devices before upload, so your photos and their metadata reach the server only in encrypted form, and its face and search features run in the apps instead. Our recommendation: pick Immich if you want a feature-rich library you can browse and back up as plain files, and pick Ente if end-to-end encryption matters most, or if you want the option of Ente's hosted service as well.

Side-by-side (from each project's docs)

ImmichEnte
EncryptionOriginal, unmodified photos and videos are stored as files in your upload folder (UPLOAD_LOCATION) on the host (backup and restore)End-to-end encrypted: files and metadata are encrypted with keys that never leave your devices unencrypted (architecture)
Phone backupOfficial mobile app with auto backup when opened, background backup and album selection (README)Official mobile and desktop apps; you point them at your server from the onboarding screen (post-install)
Face recognition and smart searchRun by the server's machine-learning containerRun on your device in the mobile or desktop app; not available in the web app at photos.ente.com (machine learning)
Minimum server hardware6 GB RAM (8 GB recommended), 2 cores (4 recommended); 4 GB works with machine learning disabled (requirements)1 GB RAM and 1 CPU core with the quickstart script (requirements)
Storage backendLocal folder on the hostS3-compatible object storage; the quickstart ships MinIO (object storage)
DatabasePostgreSQLPostgreSQL
Hosted optionSelf-hostedEnte's own cloud, with a 10 GB free plan (ente.io)
LicenceAGPL-3.0 (repo)AGPL-3.0 (repo)

Both need Docker with the docker compose plugin; Ente asks for Compose 2.30 or newer.

Hardware: why Ente's server is so small

Ente's docs say the server is a lightweight Go binary because "most of the intensive computational tasks are done on the client". That's why it can run with 1 GB of RAM. The work moves to your phone and computer instead: to build face and search indexes, the apps download your originals, analyse them on the device, then encrypt the indexes and sync them to your other devices. Ente recommends turning machine learning on in the desktop app first for a large library, since desktops index faster than phones.

Immich does the reverse. Its server runs the machine learning, so it needs more RAM (6 GB minimum), and your phone only uploads. If you have a mini PC or NAS with spare memory, that's usually the simpler setup. See Immich hardware requirements and Immich machine learning setup.

Setting up and running it

Immich: one Docker Compose stack with a .env file; follow Immich with Docker Compose.

Ente: the official quickstart script creates a my-ente folder, starts the containers, and serves the Photos web app on port 3000 and public albums on port 3002. A few things the docs ask you to do afterwards:

Backups are different

With Immich, you back up the database dump and the upload folder, and the photos in that folder are ordinary files. See Immich backup and restore.

With Ente, the docs say you need three things to get your photos back: the encrypted files (object storage), the database, which holds the per-file encryption keys, and your password. Back up the object storage and the database, plus museum.yaml and any other config. Ente's docs also recommend that people new to self-hosting keep a plain (unencrypted) export of their photos, which the Ente CLI or desktop app can automate. And if you rely on your instance backup, do a full test restore.

Moving from Google Photos

Both can take a Google Takeout export. For Immich, use Google Photos to Immich. Ente imports Takeout in its desktop app (Upload, then Google takeout) and reads the JSON metadata files that come with it. Ente recommends turning on machine learning before a large import, so photos are indexed as they upload and don't need to be downloaded again later.

Who each one suits

Our recommendations:

Related: Immich vs PhotoPrism · Nextcloud vs Immich · Immich hardware requirements

Sources

Related guides