AI face search for event photos

Every photo of you.
Found by one selfie.

Organizers upload the whole event. Guests take a single selfie and watch every shot they appear in develop into focus — powered by on-device face recognition that never keeps the selfie.

512-D
signature
1
selfie
0
stored
contact sheet · roll 01scanning…
match 0.94
match 0.89

2 matches · f/1.8 · iso 800

Selfies processed in memoryPrivate streamed galleriesGPS stripped on uploadNo public image links

The flow

From camera roll to everyone's gallery — in three moves.

01step

Upload the event

Bulk-drop the day's photos. We optimize each one, strip GPS, and index every face in the background — no tagging.

02step

Share one link

Send a QR code or link. Guests log in; an optional 4-digit PIN keeps a gallery invite-only when you need it.

03step

Selfie → develop

Each guest takes one selfie. Their matches develop into a private gallery to favorite, share, and download.

Everything in the box

Built for the person behind the camera — and everyone in front of it.

Find every face

InsightFace recognition matches a guest to every photo they're in — no manual tagging, no albums to sort.

recognition

Private by default

Galleries live behind login. Flip on a 4-digit PIN for invite-only events; owners and admins bypass it.

access

A gallery worth sharing

Full-bleed justified layout at true aspect ratios, folders to organize, and one-tap favorites.

gallery

Invite your shooters

Add collaborators so the whole crew can upload. Owners and collaborators see everything.

teamwork

One-tap downloads

Guests download a single shot or the whole set — optimized, with GPS and EXIF stripped clean.

export

Selfies never stored

The selfie is embedded in memory, matched, and discarded. No biometric template is ever kept.

privacy

The technology

Recognition, demystified.

Every face in an uploaded photo becomes a 512-dimensional signature — a list of numbers, not a picture. A guest's selfie is turned into the same kind of signature and compared by cosine similarity. Cross the calibrated threshold, and it's a match.

  • Detect. Faces are found and aligned, even on rotated phone shots and HEIC.

  • Embed. Each face → a 512-D vector. The selfie is embedded in memory only.

  • Match. Cosine similarity ranks every photo; the swappable model keeps it license-clean.

face → signature
512-D signaturecosine ≥ threshold → match

Privacy is the architecture

Biometrics you can trust because they don't stick around.

Private by design

Ephemeral selfies

A guest's selfie is embedded, matched, and dropped in the same request. It's never written to disk, logged, or turned into a stored template.

A locked vault

Photos sit in a fully private bucket with public access blocked. The app streams bytes only to a logged-in guest — no presigned or public image URL is ever exposed.

Encrypted & yours

Every connection is HTTPS; files are encrypted at rest. You choose whether originals are kept, and you can delete any photo — or the whole event — for good.

A glimpse

A gallery that develops.

Real galleries stream privately to each signed-in guest — this is a stylized preview of the contact-sheet layout.

match 0.97
match 0.91
match 0.95

In the field

Organizers stopped fielding “can you send me the photos?”

We shot 3,000 photos at the reception. Guests found themselves in seconds — I didn't tag a single face.

Marisol Vent

Wedding photographer

Attendees used to email us for their conference photos for weeks. Now it's one link and a selfie. Support tickets basically vanished.

Devan Rao

Events lead, Northwind Summit

The selfie-isn't-stored part is what got it past our parents' group. Everyone trusts it, everyone uses it.

Priya Anand

Junior football club

Load the roll

Turn your event's camera roll into everyone's favorite gallery.

Create an event, upload the photos, share one link. Guests do the rest with a selfie.