Where Immich wins
If you already run a home server, or want to, Immich is the more complete answer: a self-hosted replacement for a cloud photo library, with search as one feature among many.
- It is free AGPL software. The optional lifetime product keys ($25.00 individual, $100.00 per server) support the project; Immich says “we’re never going to force anyone to pay”.
- It backs up the phone to a server you control, so the library survives the phone.
- Its smart search is contextual CLIP search, and the admin can pick the model, including multilingual models “so users can search in their native language”.
- It searches text in images (OCR) and finds people from recognised faces.
- Its duplicates utility uses machine learning to find assets that “look visually similar”.
- It has a large community — 117,646 GitHub stars as of October 2026 — and its App Store privacy label is “Data Not Collected”.
Where Queryable adds something
Queryable is for people who will never run a server and still want to search their phone by description.
- No server, Docker, port or home network: install the app, build the index and search, all on the iPhone.
- It keeps working with no network at all once photos are indexed. Immich’s docs read for this page do not say whether the iOS app can search with the server unreachable.
- Search with a sample photo and a “Similar Photos” panel on every result; Immich’s Searching page does not document search by example image.
- A cleanup that sorts look-alikes into three tiers from one scan and pre-selects the keeper, inside the same app as search.
Compare
Every Immich cell is a fact Immich publishes — in its documentation, its blog or its US App Store listing — numbered to its source. Where Immich publishes nothing in those pages, the cell says “Not published”.
| Dimension | Immich 3.3.5 (iOS app + server) | Queryable 2.1.6 |
|---|---|---|
| Price (as of October 2026) | Free app and free server software. Optional lifetime product keys: $25.00 for an individual, $100.00 for a server. You supply and run the server hardware. [1][2] | $4.99 on the US App Store, paid once. No subscription, no in-app purchases, no free tier and no trial. [8] |
| What you need to run it | A server you run: the listing says “you will need to run/manage the server on your own in order to use the app,” and the app logs in to a server URL such as http://<machine-ip-address>:2283. [1][3] | Any iPhone or iPad running iOS 18.0 or later. No Apple Intelligence requirement, no account and no sign-in. [8][9] |
| Where processing happens | On your server’s machine-learning service; GPU acceleration for Smart Search and Facial Recognition is supported only on Linux and Windows (WSL2) servers. [1][5] | On the device. Indexing and search run through Core ML and Metal on the phone, the app makes no network request of its own, and it ships no analytics or crash-reporting SDK. [9] |
| Describe a scene in words | Yes: contextual CLIP search stored in Postgres. “The default search model is fast, but there are many other options,” including multilingual models. [4] | Yes, on every supported device. A sentence is encoded on the phone by a MobileCLIP-S2 text model into the same 512-dimension space as the photos, and the library is ranked by similarity. [9] |
| Search with an example photo | Not published in the Immich pages read for this comparison. [4] | Yes. A camera button in the search bar takes a photo or picks one from the library and uses it as the query, and every photo’s detail view has a “Similar Photos” panel. [9] |
| Text inside photos | Yes. Searchable OCR: “Optical Character Recognition (OCR): Text in images”. [4] | No. Queryable performs no text recognition of any kind; a screenshot is found by what it looks like, never by the words printed in it. [9] |
| Faces | Yes: “People: Faces that are recognized in your photos/videos.” [4] | No face, people or pet recognition, and no albums, places or map. [9] |
| Duplicates and similar photos | A machine-learning duplicates utility, on by default, detects assets that “look visually similar”; you choose which to keep and which to trash. [6] | Three tiers from one scan — Exact Duplicates, Nearly Identical and Similar Scenes — with an aesthetic score pre-selecting the photo to keep and one system confirmation dialog for the batch. Stills only: videos are never grouped or deleted. [9] |
| Search with no network | Not published in the Immich pages read for this comparison. The docs do not say whether the iOS app searches with the server unreachable. [3] | Once a photo is indexed, searching for it works with networking switched off. iCloud-only originals are stored as a placeholder and not retried, so make them available on the device in Apple Photos before building the index. [10] |
| Backup | Yes — the app’s listing subtitle is “Self-hosted Backup Tool”; photos are backed up to your own server. [1] | None. Queryable reads the system photo library and never copies photos anywhere. The index lives on one device and is not synced or backed up; two devices means indexing twice. [9] |
| Open source and privacy label | AGPL-3.0, 117,646 GitHub stars on 2026-10-30. App Store privacy label: “Data Not Collected”. [7][1] | Partly. The MIT-licensed core is public at mazzzystar/Queryable (more than 2,900 GitHub stars), but it stops at 29 March 2026: everything added in 2.1.2–2.1.6 is in a private repository. App Store privacy label: “Data Not Collected”. [11][8] |
| Result quality, accuracy and speed | Not tested. | Not tested. |
Result quality, accuracy and speed: No ranking quality, recall, precision, indexing speed or search latency was measured for this page. A claim of that kind requires the controlled hands-on test in the project’s comparison research protocol, and this page is deliberately published without one.
Which to choose
Choose Immich if…
- You run, or want to run, a home server or NAS.
- You want a full self-hosted photo library: backup to your own server, OCR and faces.
- You want to pay nothing for the software.
Choose Queryable if…
- You will never run a server and want search on the phone alone.
- You want search to keep working with no network at all.
- You want sample-photo search and a tiered similar-photo cleanup in one app.
What this page does not claim
Immich does far more than Queryable; this page compares only the parts that overlap.
- Neither app’s search quality was tested, and Immich’s results depend on the model its admin picks.
- Server hardware and running costs are not estimated here; they depend entirely on your setup.
- Cells marked “Not published” mean the Immich pages read do not document the feature, not that Immich lacks it.
Methodology
Dates and storefront: Immich facts were read on 30 October 2026 from four pages of the Immich documentation (Searching, Mobile App, Hardware-Accelerated Machine Learning, Duplicates Utility), the Immich product-keys blog post, the GitHub repository and the US App Store listing for Immich 3.3.5 (minimum iOS 15.0). Queryable is version 2.1.6, US App Store, minimum iOS 18.0.
What was tested: nothing. This is a facts-only comparison published under the Tier 1 rules of the project’s comparison research protocol: every Immich cell is a fact Immich publishes, with its source numbered; every Queryable cell describes the shipping app; anything neither side publishes says “Not published”, and result quality is marked “Not tested”. No row states a ranking, accuracy, recall or speed figure. Queryable publishes this page, so read it as our argument checked against Immich’s own pages, not as a neutral review. Corrections are welcome through the project’s GitHub repository.
Citations
Every Immich statement above was read from the pages below on the date shown. Vendor pages change without notice; if one now says something different, the vendor is right and this page is stale.
- Immich — US App Store listing (v3.3.5)
- Immich blog — Immich product keys
- Immich docs — Mobile App
- Immich docs — Searching
- Immich docs — Hardware-Accelerated Machine Learning
- Immich docs — Duplicates Utility
- GitHub API — repos/immich-app/immich
- Queryable on the US App Store Source for the one-time price, the absence of in-app purchases, the minimum iOS version and the “Data Not Collected” privacy label.
- How Queryable works The on-device model, the local index, the search path and the app’s documented limits, taken from the shipping 2.1.6 build.
- Offline AI photo search The iCloud-only boundary and what “offline” does and does not cover during indexing.
- mazzzystar/Queryable — MIT licence The public indexing and search core. Last pushed 29 March 2026, so it predates the 2.1.2–2.1.6 features (translation, search by photo, time filters, three-tier cleanup).
Questions about Immich and Queryable
Does Immich work on iPhone without a server?
No. Immich’s App Store listing says it is a client app for Immich Server and that you need to run and manage the server yourself; the app logs in to your server’s URL. Its smart search, OCR and face recognition run on that server.
Is Immich free?
Yes. Immich is free AGPL software. It sells optional lifetime product keys — $25.00 for an individual and $100.00 for a server as of October 2026 — and says it will never force anyone to pay. You provide the server.
What does Queryable need instead of a server?
Only an iPhone or iPad running iOS 18.0 or later. Queryable indexes the photo library on the device with Apple’s MobileCLIP-S2 model and searches it there, with no account and no network request of its own.
Can Queryable search text in photos or faces like Immich?
No. Queryable has no OCR and no face recognition. Immich documents both as searchable.
Can I use Immich and Queryable together?
Yes. Immich backs up the phone’s library to your server; Queryable searches whatever is still on the phone, on the phone. Neither changes how the other reads the library.