How to use SemanticVault

SemanticVault finds any image by describing it — even offline. Type "bride in a white dress" and it matches the visual content of every photo, no tags required. It is the fastest way to find that one image in a library of tens of thousands.

What SemanticVault is

SemanticVault is the search module. It indexes your library by converting every photo into a vector — a list of numbers that summarises its visual content — and lets you search with plain-language sentences instead of filenames or tags.

Search is fully local. The multilingual CLIP model runs on your machine, so there is no cloud, no upload, and no subscription fee per query.

What it is for

Once a library grows past a few thousand photos, finding anything becomes its own problem. SemanticVault exists so you never have to scroll through folders again. You describe the photo, and it appears.

What you can search

• Natural language — "sunset over mountains", "dog in snow".
• Boolean search — combine terms with AND, OR, NOT.
• Image search — drop in a reference photo and find similar ones.
• Sketch search — draw a rough shape and find matching compositions.
• Colour search — find photos by a dominant colour.
• Object detection — find and count objects like people or dogs.

How to use it

1. Index your library once (the app builds the vectors).
2. Type a description into the search field.
3. Refine with filters — rating, file type, year, camera, faces only.
4. Save recurring searches as smart collections.

Smart collections like "All RAW files" or "Favourites" are one click. Your own saved searches sit right next to them.

What it brings you

• Answer client requests in seconds instead of hours.
• Find photos by mood, subject, or scene — not just by folder.
• No manual tagging, ever.
• Everything stays private on your computer.

FAQ