Immich — High-Performance Self-Hosted Photo & Video Management

What Immich Does

Immich provides a complete photo management experience:

• Auto Backup: Automatic photo/video backup from iOS and Android with background upload
• Timeline View: Chronological photo timeline with year/month navigation and memory flashbacks
• AI Search: Search photos by describing content ("dog on beach", "birthday cake", "sunset")
• Face Recognition: Automatic face detection and grouping, name your people for easy filtering
• Shared Albums: Create shared albums with family/friends, with activity feeds and comments
• Map View: View photos on a world map based on GPS metadata
• Video Support: Full video playback including RAW video formats and live photos
• Multi-user: Multiple accounts with separate libraries and shared spaces
• External Libraries: Mount existing photo directories without moving files

Architecture

┌─────────────┐     ┌──────────────┐     ┌──────────────┐
│ Mobile App  │────▶│  Immich      │────▶│  PostgreSQL  │
│ (Flutter)   │     │  Server      │     │  + pgvector  │
├─────────────┤     │  (NestJS)    │     └──────────────┘
│ Web App     │     └──────┬───────┘
│ (SvelteKit) │            │         ┌──────────────┐
└─────────────┘     ┌──────┴───────┐ │  Redis       │
                    │ ML Service   │ │  (Job Queue) │
                    │ (Python/ONNX)│ └──────────────┘
                    └──────────────┘

Self-Hosting

Docker Compose (Official)

services:
  immich-server:
    image: ghcr.io/immich-app/immich-server:release
    volumes:
      - ${UPLOAD_LOCATION}:/usr/src/app/upload
    environment:
      DB_URL: postgresql://immich:immich@database:5432/immich
      REDIS_HOSTNAME: redis
    depends_on:
      - redis
      - database
    ports:
      - "2283:2283"

  immich-machine-learning:
    image: ghcr.io/immich-app/immich-machine-learning:release
    volumes:
      - model-cache:/cache

  redis:
    image: redis:7-alpine

  database:
    image: tensorchord/pgvecto-rs:pg16-v0.2.0
    environment:
      POSTGRES_PASSWORD: immich
      POSTGRES_USER: immich
      POSTGRES_DB: immich
    volumes:
      - pgdata:/var/lib/postgresql/data

volumes:
  pgdata:
  model-cache:

Hardware Requirements

Key Features

AI-Powered Search (CLIP)

Search: "red car in parking lot" → finds matching photos
Search: "people hiking in mountains" → finds outdoor group shots
Search: "document with text" → finds screenshots and documents

The ML service uses CLIP models for semantic image search, running entirely offline on your server.

Face Recognition

• Automatic face detection in all uploaded photos
• Cluster similar faces together
• Name people and filter your library by person
• Merge duplicate face clusters
• Hide specific faces from view

Memory Flashbacks

Like Google Photos' "On this day" feature:

• Daily notifications of photos from the same date in previous years
• Animated memory slideshows
• Shared memories with family members

Mobile App Features

• Background auto-backup (even when app is closed)
• Wi-Fi only backup option
• Selective album backup
• Partner sharing (see each other's photos)
• Offline photo access

Immich vs Alternatives

常见问题

Q: 需要 GPU 吗？ A: 不需要。ML 服务使用 ONNX Runtime 在 CPU 上运行，4核处理器即可提供良好的搜索和人脸识别体验。GPU 可以加速处理但不是必需的。

Q: 可以从 Google Photos 迁移吗？ A: 可以。使用 Google Takeout 导出照片，然后通过 immich-cli 批量上传。元数据（日期、GPS、描述）会被保留。社区也有专门的迁移工具。

Q: 照片的原始质量会保留吗？ A: 是的。Immich 保存原始文件，不会压缩或修改你的照片和视频。支持 RAW 格式（CR2、NEF、ARW 等）。

来源与致谢

• GitHub: immich-app/immich — 97.5K+ ⭐ | AGPL-3.0
• 官网: immich.app