AIPICSQR uses deep neural networks to match every guest to every photo they appear in: automatically, privately, and in seconds.
Two pipelines run in parallel. One processes your uploaded photos, the other processes each guest's selfie. When they meet, the match happens instantly.
Photo uploaded
You upload your event photos from the venue. Each photo is queued for processing immediately.
Face detection
A neural network scans every pixel of every photo to detect all faces, even partially obscured, turned, or far from the camera.
Face vectorization
Each detected face is passed through a deep convolutional network that distils it into a 512-number fingerprint called an embedding. This captures the unique geometry of that face: spacing of features, contours, proportions.
Stored in vector index
The embedding is stored in an indexed database alongside the photo it came from. This is what gets searched when a guest arrives.
Guest scans QR
The guest scans the event QR code from their phone. No app download, no account, no form to fill.
Selfie captured
A camera opens directly in the browser. The guest takes one selfie. The system needs only a single clear image of their face.
Selfie vectorized
The exact same neural network processes the selfie and produces a 512-dimension embedding of the guest's face.
Similarity search
The guest's embedding is compared against every face embedding in the event's index. Photos where the cosine similarity crosses the match threshold are returned, typically in under two seconds.
The photo pipeline runs as soon as you upload, long before any guest arrives. So when the selfie pipeline runs, all the hard work is already done. The match itself takes milliseconds.
Traditional photo sorting matches names, tags, or manual selections. We match mathematics. Mathematics does not make mistakes or get tired.
Our face encoder is a deep neural network trained on hundreds of millions of face images. It has learned that what matters is not raw pixels but the spatial relationships between facial features: the distance between your eyes relative to your nose, the curve of your jaw, the geometry of your brow. These relationships are stable across lighting changes, slight pose variations, and the passage of time.
Every face, whether in your uploaded photo or in a guest's selfie, gets converted to a vector of 512 numbers. Think of it as a coordinate in a 512-dimensional space. Faces that belong to the same person cluster tightly together in that space. Faces of different people sit far apart. No names, no labels, no stored photos. Just geometry.
To find a guest's photos, we compute the cosine similarity between their selfie embedding and every face embedding in the event. A score close to 1.0 means the same person. We apply a carefully tuned threshold: high enough to eliminate false matches, low enough to handle glasses, different angles, or a slightly tired expression at the end of a long wedding.
Guests wearing glasses or sunglasses
Side profiles and three-quarter angles
Mixed lighting: flash, candlelight, daylight
Motion blur from dancing or celebration
Faces partially behind others in groups
Different hairstyles from the selfie to the photo
Large crowds with 300+ distinct faces
Children and elderly guests equally well
Guests captured from 3 metres or from 30
This is what your guests experience from the moment they see your QR code.
A guest sees your QR code at the venue, on a stand, on screen, or on a printed card. They open their phone camera, scan it, and the event page opens instantly in their browser. No app, no account, no password.
The guest is prompted to take a selfie. They see a live camera preview in their browser with a gentle guide ring. One tap captures their face. That is all the AI needs.
Within seconds, a gallery fills with every photo from the event where that guest appears. Just theirs. Not a single stranger's face mixed in. The photos are crisp originals, not compressed previews.
The guest taps once to download their entire personal gallery as a ZIP file, or saves individual photos. They walk out of the venue with their memories already on their phone.
Most guests have never experienced this before. They expect to wait days for photos, or to scroll endlessly through a shared album hunting for their own face. When their personal gallery appears in seconds instead, and it is beautiful, the reaction is visceral.
That moment of delight is the most powerful word-of-mouth you can generate. Guests show the gallery to the people around them. People pull out phones to try it themselves. Your QR code becomes the most-talked-about thing at the event.
Guests share photos on social media immediately
The photographer becomes the conversation starter
Families send links to relatives who couldn't attend
Every scan is a branded moment with your name on it
Arjun & Priya Wedding
Match confidence
The same face recognition technology that makes delivery instant also makes it private. There is no shared album, no scrolling past strangers, no accidental exposure of other guests.
Each guest receives only the photos where their face embedding matched. The system is architecturally incapable of showing Guest A the photos of Guest B. The query is filtered at the vector level, not post-hoc.
There is no "all photos" view for guests. There is no way to browse the full event gallery. A guest lands directly in their personal results. That is the only view that exists for them.
The selfie is converted to an embedding and immediately discarded. The raw selfie image is never written to permanent storage. We keep the geometry, not the photograph.
Face embeddings are tied to a specific event. They are not cross-referenced across events or retained for building a facial recognition database. When an event closes, the embeddings are removed.
No app. No account. No data trail.
Guests do not register, do not log in, and do not leave personal information. They scan, take a selfie, get their photos, and leave. That is the entire interaction.
The face recognition engine runs across distributed AI processing infrastructure. Processing scales automatically with the size of the event.
Photos are queued across multiple processors the moment you upload. A 3,000-photo wedding does not take three times longer than a 1,000-photo corporate event. The system scales to absorb the load.
All face embeddings are computed before the first guest arrives. When guests start scanning at the venue, the results come from a pre-built index, not from running the AI on demand. Speed is constant regardless of how many guests scan simultaneously.
The moment a speech ends or music stops, dozens of guests reach for their phones at once. The system handles concurrent selfie searches without degradation. Each query is independent and non-blocking.
“A grandmother who had never used a smartphone in her life scanned the QR, took a selfie, and found 18 photos of herself. She was in tears. That’s when I knew this was something different.”
Wedding and Portrait Photographer, Chennai
Setup takes minutes. Your first event is free. And the look on your guests’ faces when their photos appear? That part is priceless.
No credit card required · Setup in minutes