Pick a folder
Choose any folder with up to 200 photos (JPEG, PNG, WebP…). Your browser reads the files directly from disk, they are never transmitted anywhere.
Free toolIn your browserNo upload
Search your photos by describing them, "sunset at the beach", "two people laughing", "a whiteboard diagram".
Free AI. No uploads. No registration. No credit card. Everything runs locally in your browser. Your photos never leave your computer.
Start on your Mac with 200 photos. Unlock PhotoCore Full to index your complete archive across folders, external drives, and NAS.
Choose any folder with up to 200 photos (JPEG, PNG, WebP…). Your browser reads the files directly from disk, they are never transmitted anywhere.
The SigLIP 2 vision encoder (the exact model used by the PhotoCore Mac app) converts each photo into a 768-dimensional embedding, locally, using your GPU via WebGPU.
Type a description in any of 100+ languages. Photos whose meaning matches your words appear instantly, ranked by relevance.
The browser tools are built for quick analysis of up to 200 photos. PhotoCore for Mac indexes your complete archive, works completely offline and keeps your AI index permanently on your computer.
Most "online photo tools" send your images to a server. This one works the other way around: the AI model (~95 MB, one-time download from the Hugging Face CDN) comes to you, and everything after that happens on your device. Verify it yourself: open your browser's DevTools → Network tab. After the model download you will see zero outgoing requests while Semantic Photo Search works through your photos. The page's Content-Security-Policy technically blocks connections to any host other than the model CDN, so your photos cannot be uploaded, even by accident.
This is the same philosophy as the PhotoCore app for Mac: your entire photo archive, organized and searchable by meaning, with zero cloud, on drives you own, even offline.
Yes. Semantic search needs no tags, filenames or metadata: the AI looks at the image content itself, so even IMG_7843.jpg is findable by describing what it shows.
Yes. Screenshots, app UIs and web captures are searchable like any photo, try "a chat conversation" or "a settings screen".
Yes. Scanned pages, whiteboards, diagrams and slides are found by describing their content, "a flowchart", "a handwritten note", "a presentation cover".
Every photo is converted into a 768-dimensional SigLIP 2 embedding, a numeric fingerprint of its meaning. Your text query is embedded into the same space, and photos are ranked by cosine similarity, the same approach as the PhotoCore app.
There is no separate OCR step: SigLIP 2 natively reads prominent text inside images to a useful degree, so searching for visible words often works. For heavy text-document workflows, the Mac app's smart rename adds descriptive filenames on top.
Google SigLIP 2 (base, patch16, 256 px), the exact model inside the PhotoCore Mac app. It understands 100+ languages.
Yes. No payment, subscription or trial is required. The browser tools are completely free, up to 200 photos per folder.
No. Just open the page and start using it. No login, no email, nothing to sign up for.
No. There is nothing to pay and nothing to unlock.
Never. The AI model is downloaded to your browser, and every photo is analyzed on your device. The page's Content-Security-Policy technically prevents any upload.
Inside your browser, using your computer's GPU (WebGPU) or CPU (WebAssembly). The model is SigLIP 2, the same one used by the PhotoCore Mac app.
No. Your photos never reach our servers, we have nothing to see. There are no image trackers and no analytics on your files.
No images are sent to our servers. Results live only in your browser; a local cache (embeddings and thumbnails) stays on your device and you can clear it anytime from your browser settings.
Once the page and the AI model are loaded, all photo analysis happens locally. The one-time model download (~95 MB) is the only thing that needs a connection.
Chrome and Edge give the best experience (WebGPU acceleration and folder access). Safari and Firefox work too, with automatic fallbacks.
The browser version is intended for quick analysis, up to 200 photos per folder. Start on your Mac with 200 photos. Unlock PhotoCore Full to index your complete archive across folders, external drives, and NAS.
Related reading: How semantic photo search works · Find photos by description · Offline photo search
These browser tools are designed for quick analysis. PhotoCore for Mac indexes your complete archive, searches 10,000 indexed photos in under 50 ms, works completely offline and keeps your AI index permanently on your computer.
Free AI. No uploads. No registration. No credit card. Everything runs locally in your browser. Your photos never leave your computer.
Download on Mac App Store