← Back to site
Try it yourself · real face matching

Find yourself in your own photos.

Add some photos from a party or trip, then take a selfie. Lens It finds every photo you're in using the same open-source models as the app, running inside your browser.

📸 At a real eventThe photographer loads the photos on their laptop. Guests join the laptop's Wi-Fi, scan a QR code and snap a selfie on their phones.
🧪 On this pageYou play both roles in one browser: first the photographer (step 1), then the guest (step 2). The face matching is exactly the same.
Nothing is uploadedWorks offline once loadedClose the tab and it's gone
0
Setup · happens once

Load the AI models

About 38 MB, downloaded once and checked against a SHA-256 fingerprint.

Starting…
1
You're the photographer

Add event photos

Select several photos at once (hold Ctrl on a laptop). Use 10–50 photos of people, ideally some with you in them and some without.

2
Now you're the guest

Take a selfie

Face the camera in good light. Use your webcam, or choose a clear photo of your face that isn't one of the event photos. Your selfie stays on this device.

🔒 Unlocks after step 1 finds at least one face.

Your selfie

How it works: YuNet finds faces and SFace turns each one into 128 numbers. Faces are compared by cosine similarity (a match is ≥ 0.363, OpenCV's recommended threshold). Both models are Apache-2.0, from opencv_zoo, and run with ONNX Runtime Web. Results can include misses, especially for side profiles, sunglasses or dark photos. Source code · Privacy · Terms · Feedback · No cookies.

A Wandercoil Product