Local-First Semantic Photo Search
Ask your own photo gallery “where was that pasta night?” in plain English — and never send a photo off your device.
🔍 Search this (sample) gallery with plain English
Sample gallery, clearly labeled — a real local-first app runs a semantic model on your phone so search works offline and your photos never leave the device.
Why people want it
Your photo gallery has tens of thousands of pictures and exactly one search bar — the date picker. Asking it “where was that pasta night with Sam” or “the beach trip from last summer” currently means three hours of scrolling. A local-first semantic search app indexes your own photos and lets you describe them in plain English, the way you actually remember them: by people, places, vibes, and food.
People want it for the privacy and the precision. Because the model runs on your device, none of your face-tagged, location-stamped memories get shipped to a cloud — the search stays local-first, works offline, and never trains on your life. And because it understands meaning rather than filenames, you can find that one sunrise photo by describing it, even though you’d never organized it.
The demo above is a tiny on-device approximation: a keyword-aware index over a sample gallery. A real app pairs that with embeddings computed locally — same idea, laptop-scale understanding.