Skip to content

Visual search

Visual search lets you find photos by describing what is in them — "beach sunset", "birthday party", "red car in snow" — instead of guessing what the file might be called.

Searching for the phrase sunset over the sea: matching photos are returned with a Vision AI badge among the match reasons, and a note says weak visual matches were hidden

How it works, briefly

During enrichment, ArchiveMind runs an image-understanding model over your indexed photos and builds a visual index from the results. When you make a visual query, your description is matched against that index. The model ships inside the app, so visual search works fully offline — nothing is downloaded at search time.

What to expect

  • It becomes available after indexing. Visual search needs the visual index, which is built after your photos are cataloged — including in the very first session, once the model has loaded (earlier versions waited for the next launch; since 1.1.3 they do not).
  • Building the index takes a while on big photo collections. Progress is shown honestly in the interface.
  • It is a similarity match, not magic. Descriptions of the overall scene work best; highly specific queries (a particular person's name, text visible in the image) are not what this feature does.

When visual search is unavailable

If the image-understanding model cannot load on your machine, ArchiveMind says so plainly and everything else — regular search, browsing, thumbnails, duplicates — keeps working. See Visual search could not start.