Granit semantic search lets you find any media by describing it in natural language, without tagging it first. Granit understands the visual content of your images.
Describe what you’re looking for, not keywords
Type “sunset over the sea,” “portrait in natural light,” or “winter mood”: Granit analyzes the meaning of your query and the content of your images to surface the closest media, even when none of those words appear in the filename or tags.
How it works
Every image in your library is turned into a multimodal vector fingerprint, where text and images share the same space of meaning. Your query is converted the same way, then compared to all your images by similarity. Results are ranked from most to least relevant.
Search that adapts to your query
A precise query automatically tightens results around the best matches; a vaguer query widens the selection. You always get a coherent list, without needless noise.
Concrete example
You’re building a portfolio page and search for “black and white photos with motion.” In a few words, Granit gathers the matching images scattered across your projects, without you having to classify them manually in advance.
Your images stay private
Search runs only on your own media, hosted in Europe. Your images are never used to train a third-party AI model: they only help you find your own work.
Frequently asked questions
Do I need to tag my media to use semantic search?
No. Granit understands the visual content of your images directly. Custom tags and metadata remain useful, but they are not required to search.
How is it different from a keyword search?
A classic search matches words against the text attached to a file. Semantic search matches meaning: it finds images visually and conceptually close to your description, even without an exact word match.
Are my images used to train an AI?
No. Your media is hosted in Europe and never used to train a model. Search stays strictly limited to your own library.