Immich Photo Server on a VPS (2026): Private, Fast, Mobile-Friendly

Immich is a modern photo server with mobile clients.

Table of Contents

1. Why Immich
2. VPS specs
3. Install overview
4. Mobile sync
5. Backups

1. Why Immich

Self-hosted, fast, and easy to use.

2. VPS specs

SSD and enough RAM for indexing.

3. Install overview

Use Docker Compose; set HTTPS and storage paths.

4. Mobile sync

Use official apps; keep network stable.

5. Backups

Back up media and DB regularly.

See /vps-hosting for plans.

Why Immich beats other self-hosted photo apps in 2026

Immich is the photo/video library most people end up on after trying PhotoPrism, Lychee, and Nextcloud Photos. It has:

• iOS / Android apps with reliable background backup.
• Face recognition, smart search (CLIP), object detection.
• Time-lapse, map view, shared albums, partner sharing.
• Multi-user with per-user storage.

Active development pace is the highest in the self-hosted photo space.

VPS sizing reality

Immich does CPU-heavy ML on import (face detection, CLIP embeddings, smart search). Backfilling 100k photos can take hours.

The DB and Redis-equivalent are tiny; ML and image preview generation dominate.

docker-compose that runs in production

name: immich

services:
  immich-server:
    image: ghcr.io/immich-app/immich-server:release
    volumes:
      - ${UPLOAD_LOCATION}:/usr/src/app/upload
      - /etc/localtime:/etc/localtime:ro
    env_file: .env
    ports:
      - "127.0.0.1:2283:2283"
    depends_on: [redis, database]
    restart: always

  immich-machine-learning:
    image: ghcr.io/immich-app/immich-machine-learning:release
    volumes:
      - model-cache:/cache
    env_file: .env
    restart: always

  redis:
    image: docker.io/redis:6.2-alpine
    healthcheck: { test: [ "CMD-SHELL", "redis-cli ping || exit 1" ] }
    restart: always

  database:
    image: ghcr.io/immich-app/postgres:14-vectorchord0.4.3-pgvectors0.2.0
    environment:
      POSTGRES_PASSWORD: ${DB_PASSWORD}
      POSTGRES_USER: ${DB_USERNAME}
      POSTGRES_DB: ${DB_DATABASE_NAME}
    volumes:
      - ${DB_DATA_LOCATION}:/var/lib/postgresql/data
    restart: always

volumes:
  model-cache:

Reverse proxy 2283 with Caddy or nginx; never expose directly.

.env that matters

UPLOAD_LOCATION=/srv/immich/library
DB_DATA_LOCATION=/srv/immich/postgres
DB_PASSWORD=CHANGE_ME
DB_USERNAME=postgres
DB_DATABASE_NAME=immich
TZ=Europe/Amsterdam
IMMICH_VERSION=release

First-run trap

After deploy, the first ML scan can saturate CPU for hours. If you notice the box being unresponsive:

# pause ML jobs from the admin UI: Administration > Jobs
# or rate-limit via concurrency

In Administration > Settings > Machine Learning, lower concurrency from 2 to 1 on smaller VPS.

Backups

Two things to back up:

1. UPLOAD_LOCATION (the photos themselves).
2. The Postgres database (metadata, faces, albums).

# postgres dump
docker exec -t immich-database pg_dumpall -c -U postgres > /backups/immich-$(date +%F).sql

If you only back up the photos, you lose facial recognition, albums, and metadata on restore.

Common breakage