Doteka uses a local CLIP neural network to semantically understand your entire gallery. Type "red car in the snow" or "birthday dinner 2023" — and get instant results. No internet required.
Doteka combines semantic AI search with OCR text recognition for truly powerful photo discovery.
Three steps from install to finding any photo in your library.
On first launch Doteka downloads the vision and text CLIP models (~300 MB total) over Wi-Fi. These run entirely on your device — no internet is needed after that.
A background worker scans every photo, runs the vision model on each image to generate a semantic embedding, and stores everything in a local database. OCR runs simultaneously to extract any text found in images.
Type any description in the search box. Doteka converts your query to an embedding using the text model, computes cosine similarity against all photo embeddings, and shows the most relevant results instantly.
Doteka was built with privacy as the foundation, not an afterthought. We have no servers that receive your photos, no analytics that track your searches, and no account system that ties data to your identity.