What Is PhotoPrism?
PhotoPrism is a self-hosted web application for browsing, organizing and searching photos and videos on a server you control. Its website calls it "an AI-powered, privacy-first app"; the code is on GitHub under the AGPL, and the company says it is self-funded and independent.
The free Community edition is the full open-source release. Paid memberships and team licences unlock extra features, and the project lists a managed host as an alternative to running the server yourself.
Get It
PhotoPrism is a server you install yourself; the recommended route is Docker Compose, and there is no one-click desktop or phone installer.
Option | How to get it |
|---|---|
| Docker Compose | Setup guide: download a config file, edit it, run one command |
| Windows helper | curl.exe and install.bat fetch the config and start the server, per the same guide |
| Linux packages | tar.gz packages for experienced users; manual configuration |
| Managed hosting | PikaPods, from $6.50 per month per the pricing page |
| Community edition | Free, AGPL-3.0, self-hosted; unlimited storage and instances |
| Memberships | Essentials €2.00, Plus from €6.00 per month, shown in euros on the pricing page |
| Website | photoprism.app |
| Privacy policy | photoprism.app/privacy |
| Source code | github.com/photoprism/photoprism |
| Public demo | try.photoprism.app; the project asks you not to upload private pictures |
Version reviewed: stable build 261007-65faaae5d, published on 7 October 2026 per the GitHub release, with commits on the development branch through 24 October 2026. The release ships only Linux amd64 and arm64 tar.gz files, and the project steers most users to its Docker image. The release notes also list a development preview, which is not the version reviewed. Prices are the figures shown on the pricing page; the managed host is priced in US dollars.
Requirements
The docs ask for at least 2 CPU cores, 3 GB of RAM and a 64-bit operating system, plus swap and SSD storage for large libraries.
Server
- What the sources say:
- 2 cores and 3 GB RAM minimum; RAM in GB should match the core count
Swap
- What the sources say:
- At least 4 GB recommended; RAW files and panoramas may need more
Runtime
- What the sources say:
- Docker or Podman on Linux, macOS or Windows; amd64 and ARM64 images, including Raspberry Pi
Database
- What the sources say:
- SQLite or MariaDB 10.5.12+; the docs call SQLite a poor fit when performance matters
GPU
- What the sources say:
- Optional; NVIDIA cards can run the built-in models, using a separate cuda image
Internet
- What the sources say:
- Needed for reverse geocoding and world maps; the docs ask firewalls to allow the geocoding API
Browser
- What the sources say:
- A current browser; maps need WebGL 2
These minimums come from the project's docs; PromptQuorum did not measure indexing time or memory use on any hardware.
How to Get Started
These steps follow the project's Docker Compose guide; PromptQuorum has not run them.
- 1Install Docker
Why it matters: Install Docker (or Podman) on the machine that will hold your photos; the guide covers Linux, macOS and Windows. - 2Download and edit the compose file
Why it matters: Take the compose.yaml for your platform from the setup guide, then set an admin password and the folders that hold your pictures. - 3Start the server
Why it matters: Run docker compose up -d in that folder, open the server's address on port 2342 in a browser and sign in as admin. - 4Index your library
Why it matters: Add or import your pictures and let the first indexing run; the project's First Steps tutorial explains the settings that control it.
Key Features
Every item below comes from the project's website, README or documentation; none has been tested by PromptQuorum.
- Search and browsing. Filters that can be combined cover labels, location, resolution, colour and quality; views include People, Places, Moments, Calendar, Labels and Folders.
- Formats. RAW images, Live Photos, Motion Photos, videos and 360° panoramas are indexed, and duplicates are detected.
- Metadata. Exif, XMP, ExifTool JSON and Google Photos data can be read and merged.
- Access. An installable web app for phones, WebDAV access from Windows Explorer or Finder, and OpenID Connect single sign-on.
AI Features and Where They Run
The built-in models run on your own server through ONNX; only the optional OpenAI engine sends pictures to a cloud service.
- Labels and adult-content detection. A built-in ONNX model labels pictures and a second one flags NSFW content; the 7 October 2026 release replaced the earlier TensorFlow models and no longer supports custom TensorFlow models.
- Faces. YuNet locates faces and SFace turns each into a vector for grouping; older libraries keep the FaceNet model until migrated, and the release says new clusters are now named automatically where possible.
- Captions and richer labels. Optional: an Ollama server you run yourself, or the OpenAI API in the cloud with your own key, can generate captions and labels.
- GPU. The built-in models can use an NVIDIA card; for Ollama label generation the project reports 1 to 4 seconds per image on an RTX 4060, and 10 seconds to over a minute without a GPU.
- Switching off. Setting the labels model to "none" disables label generation, while face detection is turned off with a separate setting.
Data Flow and Privacy
The privacy policy, last updated on 19 August 2026, says data that never leaves your private network cannot be collected by anyone, including the company. For a self-hosted server that means photos, albums and AI results stay on your hardware unless you send them elsewhere, for example through the OpenAI engine.
According to the policy and docs, three kinds of traffic reach outside servers. Reverse geocoding and interactive world maps retrieve data from PhotoPrism and from MapTiler AG in Switzerland, and a geocoding request carries the location being looked up plus your server's public IP address. Instances with a membership key also contact the project to renew it, sending a random instance ID, the software version, operating system type and CPU architecture. Finally, a server reachable from the internet must sit behind an HTTPS reverse proxy, otherwise passwords and files travel in clear text.
📌Note: These are the project's declarations, not audit results. PromptQuorum has not inspected network traffic or reviewed the source code beyond the repository metadata. The policy names PhotoPrism UG as the operator of its own services and links a GDPR compliance statement.
Who Should Use It
- Self-hosters with a Docker server or NAS who want a private alternative to cloud photo services.
- Photographers with large RAW and video libraries who need search by label, place and face.
- Households that want one shared library reachable by browser on every device.
What We Could Not Verify
- AI quality. PromptQuorum did not test label accuracy, face clustering or search relevance; the speed figures above are the project's.
- Phone experience. The project ships a web app only; the quality and terms of the third-party mobile apps in its docs were not checked.
- Edition boundaries. Which features sit behind a paid membership can change between releases; check the pricing table before relying on maps or extra user roles.
- Licence details. GitHub's detector reports "Other", while the README and docs state AGPL-3.0 for the Community edition; the terms of the paid editions were not reviewed.
- Resource use. The docs say indexing large RAW libraries can exceed the minimum memory; real-world usage was not measured.
Competitors and Alternatives
The directory has one other image-library manager with a review, plus a self-hosted app that tags a different kind of content.
PixlStash
- Platforms:
- Windows, macOS, Linux
- Price / license:
- Free / GPL-3.0
- Key difference:
- Desktop app with local tagging, face and semantic search; PhotoPrism is a browser-served server
Articles about PixlStash (1)
- PixlStash Review: Self-Hosted Image Library Manager with Local AI TaggingUpdated October 17, 2026
Karakeep
- Platforms:
- Web, iOS, Android
- Price / license:
- Free / AGPL-3.0
- Key difference:
- Bookmark and notes keeper with AI tagging, not a photo library
Articles about Karakeep (3)
- Karakeep Review: Self-Hosted Bookmarks with AI Auto-TaggingUpdated September 18, 2026
- Local Knowledge & Retrieval Tools Compared (2026): RAG, Document Chat, Vector Databases, Notes and SearchUpdated September 20, 2026
- Blinko Review: Self-Hosted AI Notes with RAG SearchUpdated September 19, 2026
Also mentioned in:
- Linkwarden Review: Self-Hosted Bookmarks with Local AI TaggingUpdated October 24, 2026
- Surf Review: Deta's Open-Source AI NotebookUpdated September 19, 2026
No other photo-library server has a PromptQuorum review yet, so these are the nearest listed tools; confirm each app's current price, license and platforms on its own page.
Frequently Asked Questions
Does PhotoPrism have an app for iPhone or Android?
The project itself offers an installable web app; its documentation lists apps from other developers, among them PhotoSync for background uploads from phones.
How do I update PhotoPrism?
With Docker Compose you pull the new image and restart the services; changing the image tag from latest to preview gives early access to development builds. Release notes can contain upgrade steps, for example the label re-run after the October 2026 AI change.
Can I move over from Google Photos?
Metadata from Google Photos can be merged, and the docs include a cloud migration guide; PromptQuorum did not try a migration.
Where does PhotoPrism keep my files?
The example config defines three folders: originals for your pictures, storage for internal files such as configuration, and import for files to be added. It also starts a MariaDB container for the database.
Is there a business version?
Yes: the Teams page lists licences from €79 per month for Starter, from €219 for Business and from €1,499 for Enterprise, with support levels that differ by tier.
Verdict
PhotoPrism is a long-running, actively updated way to keep a searchable photo library on hardware you control, with face recognition and AI labels that run on your own server by default. Against that, setup needs Docker and a config file, the project lists no native desktop or phone app of its own, place lookups and maps contact outside servers, and none of its AI quality or speed claims could be checked here. It suits people who already run services at home or on a NAS and can try the free Community edition first; anyone who wants a double-click app should look elsewhere.
Sources
- PhotoPrism website — product overview, pricing and editions, checked 24 October 2026.
- PhotoPrism documentation — setup, AI models, release notes and mobile apps, checked 24 October 2026.
- PhotoPrism privacy policy — data collection and self-hosting statements, last updated 19 August 2026.
- PhotoPrism on GitHub — README, releases and commit activity, checked 24 October 2026.