AI looks at every image or video frame and identifies what is shown: people, objects, places, text, scenes and visual styles. Each file is assigned a “visual signature” that captures its content, without relying on any existing subject headings or descriptions.
When someone searches, their text query (or, in some setups, an example image) is also turned into a visual signature. The system then compares this signature to all signatures in the collection and returns the closest matches.
Visual search is for exploration and pattern-finding. It makes an entire collection explorable, which is essential when item-level descriptive metadata is sparse or missing. In well-described collections, it offers a different way of seeing the material: researchers can switch to image-based search when their question is visual, then return to the catalogue for creators, series, dates and context. The two modes work together: find by content, understand by metadata.
More than 200 hours of historical Swiss newsreels from 1940–1975 are now searchable by visual content. Researchers can look for specific scenes or on‑screen text across the entire run of films, instead of scanning footage manually or relying only on catalogue descriptions.
Visual search is being used to unlock tens of thousands of Lethbridge Herald press photographs that were scanned but had little or no metadata. This has turned a previously inaccessible collection into a searchable online resource.
Around 40,000 digitised images from the museum’s photographic collections can now be explored by visual content. Staff and visitors can look for scenes, objects or activities across the whole set, even where no detailed descriptions exist.
The Bern City Archives used visual search to open up around 8,000 digitised photographs from their collections. Users can search by what appears in the image and then use the existing online catalogue for full description and context.
AI visual search is particularly powerful for press-agency archives, photography sets, poster and postcard collections, public-sector image holdings, and corporate image or video archives - anywhere collections are large and item-level description is limited. If you're interested in how our clients are using it, explore our case studies.
