Search with a photo.

Show it a room, a desk or a shoe. It finds each item in the photo and matches it against 61,000 Amazon products. The models run in your browser, so your photo never leaves your device.

Drop a photo here

or paste one with Ctrl+V / ⌘V

Or try one of these

Find items

YOLOS-tiny, a small object detector, runs in your browser and boxes each chair, vase or bag it sees.

Turn each item into numbers

SigLIP 2, also in your browser, turns each box into a list of 768 numbers (a vector). Similar-looking things get similar vectors.

Search all items at once

One batch request asks Qdrant for the closest products to every item. Each product has up to 4 photos stored as one multivector, so a side view can match a side view.

Steer it with words

Tap “+ in blue” or “+ wooden”. Each chip is a SigLIP 2 text vector, computed ahead of time. Qdrant re-scores the look-alikes with it: image score + 4 × text score.