When image search is better than text
Image-to-image search is useful when a scene is easy to recognize but awkward to describe. A distinctive interior, a recurring trail, a particular product, an outfit, a document layout or a style of illustration can carry more signal in pixels than in a short phrase.
It can also surface a sequence captured around the same subject even when the images are not exact duplicates. That makes it useful for finding alternate compositions before choosing the best shot.
- Other photos from a location
- Repeated products or objects
- Similar screenshots and documents
- Alternate shots of the same scene
What happens on-device
Queryable passes the selected image through its local image encoder to create a 512-value embedding. The app normalizes that vector and compares it with the saved local embeddings for your library. Its current search path uses Metal-backed matrix operations to rank a large candidate set efficiently.
The output is a similarity score, not an exact duplicate declaration. Two photos can rank closely because they share composition, colors, objects or scene type even if they were taken years apart.
Similar search versus duplicate cleanup
A visually similar photo is not necessarily redundant. A burst may contain small but important changes in expression or focus. A recurring view from the same window may be meaningful across seasons. Treat similarity as a review queue, not an automatic deletion decision.
For cleanup, use a stricter duplicate or similar-photo workflow, inspect full-resolution candidates, and preserve the version with the quality, metadata and edit state you want.
Practical note: Never use visual similarity alone as proof that two files are identical. Review before deleting.
Private reverse image search
Public reverse-image engines answer a different question: where an image or similar content appears on the web. Queryable answers which items in your personal photo library look related. It does not upload the reference image to a web index for this search.
If your goal is finding an original publisher, copyright source or online copy, use a public web image-search service instead. If your goal is reconnecting one personal image with related moments, local library search is the more relevant tool.
Common questions
Is search by photo the same as reverse image search?
The interaction is similar, but Queryable searches only your local photo library. It does not search public websites.
Will it find exact duplicates?
It can find visually close images, while the dedicated duplicate and similar-photo tools are better suited to cleanup. Always review candidates.
Is the reference image uploaded?
Queryable encodes the reference image on-device for local library matching; it is not sent to a Queryable image-search server.