Search by reference photo
Pick one image and find visually related places, objects, outfits or repeated scenes.
Private, on-device photo search
Find the photo you remember. Describe it.
Queryable turns a sentence—or another image—into a private search across your iPhone photo library. The core index and search stay on your device.
Open source2,900+ GitHub starsNo subscription
Try a memory, not a filename
a white kitten on a park lounger
2.1.2Current release
Since 2022Indie-built
✦a white kitten on a park lounger✦dark night room with a lamp✦pink sky with clouds✦a white kitten on a park lounger✦dark night room with a lamp✦pink sky with clouds
01 / Seen, not staged
Four searches from the first working app.
These are real results from a personal photo library, not a stock-photo mockup. The prompts move from an object to a color, a scene and even a feeling.
Original product recording from 2022 · 00:29 · no cloud search
02 / The 29-second test
Watch the search, not a sales reel.
The original screen recording moves through five plain-language searches without leaving the iPhone. The interface shown is the early Queryable design; the current App Store release has since added date filters, reference-image search and library tools.
- 01
Index the library once
An image encoder turns locally available photo data into compact 512-dimensional embeddings and saves them on the device.
- 02
Describe a memory
The text encoder maps a phrase into the same visual space. Non-English queries can also use local translation before ranking.
- 03
Rank locally
A GPU-backed matrix comparison scores the query against the local index and returns the closest matches.
03 / What people remember
Color, emotion, odd ideas—and offline.
Queryable is useful when the clue in your head is visual but imprecise. These are the product’s actual App Store screenshots, built from searches in the developer’s own photo library.
04 / Under the surface
Two encoders, one shared visual space.
The library is encoded ahead of time. A search only needs to encode the new phrase and compare that vector with the saved photo vectors, which is why repeated searches can stay fast and local.
A note on provenanceThe diagram is intentionally historical evidence, not a claim that every implementation detail stayed frozen since 2022.
05 / The app today
The experiment became a library tool.
The public project and current app go beyond the first text box. The following capabilities are present in the current codebase or App Store release, not speculative roadmap items.
Date-aware search
Combine a visual description with a time range when you remember roughly when it happened.
Duplicate grouping
Review similar-photo groups with an adjustable threshold instead of blindly deleting everything.
Batch actions
Select multiple results to share or delete; original-resolution exports are supported.
Live Photo and video
Open, play and share more than still images from the results view.
Local language help
On supported systems, Apple’s Translation framework can expand a non-English query locally before the results are merged.
06 / Why it exists
A personal tool that became a public project.
Ke Fang built the first working version in two weeks at the end of 2022 to answer a narrow question: could an iPhone find a remembered scene by meaning, without sending the photo library to a server?
The code was published under the MIT license so the privacy and engineering claims could be inspected. The Chinese edition is named 寻隐—literally, to seek what is hidden—a name that fits both the classical poem it references and the act of recovering a half-remembered photograph.
This page keeps that history short. The two field notes below document the original engineering choices and the longer product journey in the author’s own words.
Ke FangCreator of Queryable · independent iOS developer
Local index On your iPhone 07 / Privacy
Your library is not an AI training dataset
Queryable’s core indexing and semantic search run on your iPhone. Its index request stays offline; make cloud-only items available through Apple Photos first. Queryable does not operate a photo-upload server.
- On-device semantic index
- No Queryable photo-upload server
- Works offline after local indexing
08 / Open source
Inspect the code, not just the promise
Queryable is an MIT-licensed open-source app built around Apple’s MobileCLIP model. Developers can review how local photo embeddings and text-to-image matching are implemented.
View Queryable on GitHubSearchEngine.swift
let query = "a white kitten on a park lounger"
let textEmbedding = model.encode(query)
let matches = photoIndex
.ranked(by: textEmbedding)
.prefix(50)
// Your photos stay on this device.09 / Honest limits
Semantic search is useful, not magical.
Queryable ranks visual similarity. It does not know the personal truth behind every photograph, and the first result will not always be the one you meant.
Meaning is approximate
Abstract prompts such as an emotion are subjective. Treat the result list as a strong place to start, not a factual classification.
It is not face identification
Queryable searches visual meaning and similarity; it is not presented as a named-person or biometric recognition system.
Local data must be available
Photos stored only in iCloud need to be made available through Apple Photos before offline indexing can use them.
The first index takes work
Building a large local index uses time, battery and storage. Newer devices finish sooner; later searches reuse the saved index.
10 / Facts
A small app with verifiable details
Version, system requirement, rating and price come from the live App Store listing, checked 2026-07-14.
2.1.2Current version
iOS 18.0+System requirement
4.7 ★81 App Store ratings
$4.99One-time purchase · local prices vary
11 / FAQ
Questions, answered
01What is Queryable?+
Queryable is an iPhone app that lets you search your photo library with natural-language descriptions or a reference photo. Its core AI index and search run on-device.
02Does Queryable upload my photos?+
No photo library is uploaded to a Queryable server for indexing or search. Make cloud-only items local through Apple Photos before Queryable’s network-disabled index request.
03Does it work offline?+
After usable image data and the on-device index are available, semantic search works without an internet connection. Prepare cloud-only items through Apple Photos first.
04What iPhone version does it require?+
The current App Store release requires iOS 18.0 or later. A newer device is recommended for faster initial indexing.
05Is Queryable free?+
Queryable is currently a one-time paid download, not a subscription. The US App Store price was $4.99 when this page was updated; regional pricing can differ.
06What is 寻隐?+
寻隐 is the separate Chinese edition of Queryable, with Chinese App Store metadata and localization for users in mainland China.
Stop scrolling. Start describing.
Build the local index once, then search years of photos in the language you naturally use.
Get Queryable for iPhone