Local AI processing • No upload • Privacy-first photo search

AI Semantic Photo Search

Search through your local photo collection using natural language descriptions. Runs entirely in your browser with AI acceleration. No uploads required.

What this semantic photo search tool does

This browser-based tool lets you search through your local photo collection using natural language descriptions. It understands visual content and matches photos to your text queries — all without uploading a single image to any server.

How to search photos by description

  1. 1.Select a local folder containing your photo collection
  2. 2.AI automatically indexes each photo, extracting visual feature vectors
  3. 3.Type a description like 'sunset beach' or 'cat sleeping' to search
  4. 4.Browse results ranked by visual similarity to your query

Why use local semantic photo search

All processing happens in your browser — photos never leave your device

Understands visual concepts, not just filenames — search by describing what you see

Indexed vectors stored in IndexedDB for instant re-search without re-indexing

WebGPU acceleration with fp16 for fast indexing of large photo collections

Semantic photo search FAQ

Are my photos uploaded to a server?

No. The entire workflow runs locally in your browser. Photos are read directly from your local folder, processed on your device, and feature vectors are stored in your browser's IndexedDB. Nothing is uploaded.

Which image formats are supported?

The tool supports common web image formats including JPEG, PNG, WebP, GIF, and BMP. HEIC and RAW formats are not currently supported.

Why does it need to download data on first use?

First run downloads the AI recognition model (~350MB) to your browser cache. Subsequent visits use the cached version. The model enables understanding of visual content and text descriptions.

What if my browser doesn't support WebGPU?

The tool will fall back to CPU mode, which is slower but still functional. For the best experience with large photo collections, use Chrome 113+ or Edge 113+ which support WebGPU.