← Back to Project List
Immich 是全球最受欢迎的开源自托管照片和视频管理方案,106K+ GitHub Stars,支持移动端/Web 端全平台覆盖,具备人脸识别、CLIP 语义搜索、多用户管理、OAuth 集成等企业级能力,是 Google Photos 和 iCloud Photos 的私有化替代方案。

1. Project Overview

1.1 basic information

DimensionDescription
Project NameImmich
Organization/Authorimmich-app (Community Driven, Core Team Maintenance)
GitHubhttps://github.com/immich-app/immich
official websitehttps://immich.app
Open Source ProtocolAGPL-3.0
TypeScript(52.6 percent), Dart(28.6 percent), Svelte(13.3 percent), Python(1.6 percent)
Star number106,738
Cumulative Submissions10,378 +
Number of Release299 (latest v2.7.5,v3.0 RC in progress)
demo environmenthttps://demo.immich.app (account demo@immich.app / demo)

1.2 Project Positioning

Immich is a high-performance, self-hosted photo and video management platform that aims to be an open-source, private alternative to Google Photos / iCloud Photos. Users can deploy Immich on their own servers to get an experience comparable to that of cloud albums, while fully controlling data sovereignty.

Core Value Proposition: Privacy-first users are reluctant to upload their personal photos to the cloud, but are reluctant to give up modern photo album features such as intelligent search, face recognition, and automatic backup. Immich fills this gap in the market.

1.3 development process

-2022: Project launch, rapid accumulation of early users

-2023: Release of ML pipeline (face recognition, CLIP semantic search), functional benchmarking Google Photos

-2024: License switch from MIT to AGPL-3.0, start commercialization (Immich product key, peripheral mall)

-2025: Star Breaks 50K and Becomes the No.1 Project in GitHub Photo Management Category

-2026 (current):Star breaks through 106K,v3.0 is under development, and the plug-in system will be officially released soon.

2. Core Competence

2.1 automatic backup and synchronization

-Mobile App(iOS / Android) supports automatic backup in the background

-Deduplication mechanism: The same file will not be uploaded repeatedly, based on hash verification.

-Selective Backup: You can select the backup range by album dimension. This is not a mandatory global backup.

- Background Backup (Background Backup) : The Android side supports background insensitive backup, and the iOS side needs to run in the foreground due to system restrictions.

2.2 intelligent retrieval

Immich has built-in multi-layer retrieval capabilities, far beyond the traditional file manager:

Search MethodDescription
EXIF metadata searchaccurate screening by shooting time, device model, GPS coordinates, aperture, ISO, etc.
Face recognition and clusteringAutomatically detect faces in photos, group and cluster by people, and support manual labeling and merging
CLIP semantic searchNatural language image search based on OpenAI CLIP model, enter "dog on the beach" to find the corresponding photo
Global Map ViewGenerate thermal map based on photo GPS coordinates, support browsing photos by location
Object detectionML model automatically annotates object classes in photos

2.3 multi-user and enterprise capabilities

-Multi-user support: Support family members or organization members to have independent accounts, photo data isolated from each other

-Shared Albums: Users can create shared albums and invite specified users to upload and browse them collaboratively.

-Partner Sharing (Partner Sharing): Selective sharing of photo libraries between family members

-OAuth authentication: supports OAuth 2.0 integration (OIDC) and can be connected to enterprise SSO systems

-API Keys: provides API key system and supports third-party application integration

-Management background:Web provides administrator functions such as user management, storage quota, and system monitoring.

2.4 multimedia processing

-RAW format support: Supports direct preview and transcoding of mainstream camera RAW formats

-LivePhoto/MotionPhoto: Supports backup and playback of Apple Live Photo and Android Motion Photo

- 360 panoramic photos :Web supports interactive viewing of 360-degree panoramic photos

-Video playback: Built-in video player, support streaming and thumbnail preview

-Stack Photos (Stacked Photos): Support automatic stacking of continuous shooting photos to reduce browsing clutter

2.5 user customization

-Custom storage structure: allows users to customize the file storage path by year/month/day or other rules.

-Read-Only Gallery: Supports mounting an external repository and setting it to read-only mode without modifying the original file

-Folder View: Supports traditional folder tree navigation in addition to timeline view

-Archives & Collections: Supports photo archiving and collection tagging

-Tag System: supports manual tag management on the Web

3. Technical Architecture

Overall 3.1 Architecture

Immich uses a microservice architecture and Compose one-click deployment through Docker:

┌─────────────────────────────────────────────┐
│  immich-web(Web 前端)                       │
│  技术栈:Svelte + TypeScript                   │
├─────────────────────────────────────────────┤
│  immich-server(后端核心)                      │
│  技术栈:NestJS(Node.js)+ TypeScript          │
│  功能:API 服务、认证、权限、文件管理              │
├─────────────────────────────────────────────┤
│  immich-machine-learning(ML 服务)            │
│  技术栈:Python + ONNX Runtime                 │
│  功能:人脸检测、CLIP 嵌入、对象检测               │
├─────────────────────────────────────────────┤
│  数据层                                       │
│  PostgreSQL(元数据)+ Redis(缓存/队列)          │
│  + 文件系统(照片/视频原文件)                     │
└─────────────────────────────────────────────┘

3.2 Key Components Details

immich-server(NestJS):

-Based on the NestJS framework, TypeScript full-stack type security

-Support RESTful API and WebSocket real-time communication

-Built-in job queue system, asynchronous processing thumbnail generation, video transcoding, ML inference and other time-consuming tasks

immich-machine-learning(Python):

-Model inference with ONNX Runtime, CPU and GPU support (CUDA / OpenVINO)

-Face detection: based on MTCNN or RetinaFace

-Face recognition: based on ArcFace feature extraction

-CLIP Semantic Search: OpenAI CLIP ViT-B/32 model

-All ML inference is performed locally, photo data does not come out of the server

Mobile Applications:

-Flutter(Dart) cross-platform development, iOS and Android unified code base

-Support for native background tasks (Android WorkManager / iOS BGTaskScheduler)

-Use Isar database for local cache and offline storage

3.3 deployment method

-Docker Compose (recommended): stand-alone deployment, suitable for home users and small organizations

-Kubernetes: supports Helm Chart deployment, suitable for enterprise-level multi-node expansion

-Unraid/TrueNAS: Community provides one-click installation templates

- Hardware requirements are not high : Raspberry Pi 4(8GB) can run basic functions; If ML inference is required, NVIDIA GPU or Intel OpenVINO is recommended.

3.4 plug-in system (v3.0 key features)

v3.0 introduces a plug-in system that allows developers to write feature extensions for Immich:

-Plug-in Type: Supports metadata extractor, storage template provider, webhook action, and other types.

-TypeScript SDK: provides a type-safe plug-in development SDK.

-Independent deployment: The plug-in runs as a separate process and communicates with Immich Server through API

- Market Potential : Community ecology similar to Home Assistant, third-party developers can contribute plug-ins and make profits.

4. Market positioning and competition analysis

4.1 Competitive Landscape

ProductTypeCostData SovereigntyAI CapacityMulti-User
Google PhotosCloud SaaSFree 15GB/Paid ExpansionNoneStrongLimited (Partner Sharing)
iCloud PhotosCloud SaaSFree 5GB/Paid ExpansionNoneStrong (End Side)Limited (Home Sharing)
Synology PhotosNAS KitSynology NAS requiredAvailableMediumAvailable
PhotoPrismOpen Source Self-HostingFree/Paid EditionHaveMediumNone (Personal Edition)
ImmichOpen Source Self-HostingFree/Optional DonationHaveStrongHave
Nextcloud MemoriesOpen source self-hostingFreeWithWeakDependency Nextcloud

4.2 Immich's Differentiation Advantage

  1. The only open source solution with" Google Photos Class AI Experience "," Full Data Sovereignty "and" Multi-User Support "
  2. Photoprism lack native mobile App and multi-user support; Synology Photos binding hardware
  3. The mobile experience is highly consistent with the cloud photo album, reducing the user switching threshold
  4. The community size of 106K Stars means long-term maintenance guarantee and rich third-party integration

4.3 target user group

User GroupsScenariosWillingness to Pay
Tech Enthusiast/Individual DeveloperFamily Photo Manager, NAS UserMedium (Donate/Peripheral)
Small Business/StudioInternal Asset Management, Collaborative SharingHigh (Product Key/Business License)
Privacy Sensitive Industries (Lawyers/Doctors/Finance)Customer Data Compliance StorageHigh (Business Authorization Customization Services)
Educational Institutions/Nonprofit OrganizationsEvent Photo Archiving and SharingMedium
NAS VendorPre-installed Value Added ServicesHigh (OEM Authorization)

5. Business Model

Immich uses Open Core peripheral services business model:

  1. Immich Product Key (Product Key): Voluntary purchase, unlock premium feature priority and support (Obsidian-like Catalyst License)
  2. peripheral mall (immich.store) : brand peripheral goods
  3. Commercial Licensing: The AGPL-3.0 has Copyleft restrictions on SaaS deployment, and enterprises need to purchase a commercial license if they need closed-source integration or privatization customization.
  4. Potential OEM Cooperation: Pre-installed with NAS vendors (Synology, QNAP, UGREEN, etc.) and realized through license fees or revenue sharing

Reference project: Gitea → Forgejo, Grafana → Grafana Enterprise open source commercialization path.

6. Pre-sales entry point

6.1 Why do customers need Immich?

Pain Point 1: Privacy Anxiety of Cloud Albums

Google Photos and iCloud scan user photos for AI training and advertising purposes (although Google states that they are not used for advertising). For privacy-sensitive users and organizations, this is unacceptable. Immich has all the data on the client's own server, and AI inference is performed locally.

  • Pain Point 2:NAS's own album function is weak *

Most NAS (Qunhui, Wei Unicom, etc.) come with limited photo management applications: lack of face recognition, semantic search, poor mobile experience. Immich can be deployed separately as a value-added application for NAS.

Pain Point 3:Google Photos Free Lunch Ends

Google canceled unlimited high-quality photo storage in 2021, and a large number of users were forced to pay or find alternatives. The one-time hardware investment for self-hosted Immich is much lower than the cost of a long-term cloud subscription.

6.2 Pre-sales Speech

-To individual users:"Are you willing to give Google 20 years of photos of your family? Immich lets you use Google Photos experience, but the data is only on your own hard drive."

-For small businesses:"Customer profile photos have compliance risks on the public cloud. Immich can be deployed on your intranet server to meet GDPR/Personal Information Protection Act requirements."

-To NAS vendors:"Your users need a better photo management experience than Synology Photos. Integrating Immich can add 30 percent to your hardware."

7. Risks and Challenges

RisksDescriptionsMitigations
Destructive UpdateData Structure Changes May Exist in Major Version Upgrade (v2 → v3)Official Provides Detailed Migration Guide and Database Backup Suggestions
Deployment thresholdNon-technical users are not familiar with DockerCommunity provides Unraid/TrueNAS one-click template; One-click installation script can be introduced in the future
Performance bottleneckML inference requires GPU, and the experience of low-end devices such as strawberry pie is poorThe architecture supports plug-in independent ML inference nodes
Competition pressureCloud vendors may introduce more aggressive pricing strategiesData sovereignty and privacy are irreplaceable core advantages
Immature commercializationRelies on donations and peripherals, limited revenue scaleReferring to the Grafana path, enterprise licensing is a more sustainable model

8. One sentence conclusion

  • * Immich is the best choice for Google Photos' open source replacement. 106K Stars and mature mobile experience make it not only a technical toy, but also a "photo management privatization infrastructure" with commercial conditions ". For privacy-sensitive users, NAS vendors and small and medium-sized enterprises, Immich fills the gap in the market of "smart experience and data sovereignty. **<| end_of_thinking |>

Step 4: The analysis report has been written (8 chapters, immich-app/immich). Now register to gen.py and generate the HTML.

<||DSML||tool_calls>

<||DSML||invoke name="TodoWrite">

<||DSML||parameter name="merge" string="false">false