What Facial Recognition Means Here.

Editor's note: This is part of VisageAir Explained, a series describing who we are, what we make, and the principles that guide our work.

"Facial recognition" is two words that carry a lot of baggage, and we use them anyway, because they are the accurate name for the technology and we would rather explain it than hide behind a softer word. So here is what it is, what it can do on the phone in your pocket, what VisageAir™ builds on top of it, and the one idea that makes all of it worth doing: you, measured against you, over time.

One framing shapes everything that follows. VisageAir produces general wellness and longevity information. It is not a medical device, and it is not intended to diagnose, treat, cure, or prevent any disease. Our pilot opens December 2026. What you read below is the standard we are building to, stated plainly so you can hold us to it.

What facial recognition is

Facial recognition is software that finds a face in an image, locates its features, and builds a map of its structure: the corners of the eyes, the bridge and tip of the nose, the line of the jaw, the edge of the lips, the point of the chin, and hundreds of points in between. That map is a geometry. It says where things are and how they sit relative to each other.

Once you have that map, there are two very different things you can do with it. You can compare it against a database of other faces to answer the question "who is this?" That is identity matching, and it is what people picture when they hear the words. Or you can compare it against earlier maps of the same face to answer a different question: "how has this face changed?" That is analysis, and it is the only thing VisageAir does. Same technology, opposite job. We do not match you to anyone. We match you to you.

What your phone already sees

The remarkable part is how much of this a modern phone does before you ever press the shutter. The camera system finds your face in real time. It places several hundred landmark points on it, live, dozens of times a second. It knows the angle of your head in three axes, how far you are from the lens, how even the light is, and whether your expression is neutral. On phones with a depth sensor, it measures the surface of your face in real millimeters, not just pixels, so a distance across the jaw is a true distance.

Most apps throw all of that away and just take a picture. VisageAir is built to use every bit of it. The guided frame is not decoration. It reads the live landmarks and pose, and it will not fire until your face sits at the same distance, the same angle, in even light, with a neutral expression, as your last capture. Then it takes a short burst and keeps the sharpest, best-aligned frame. Consistency at the moment of capture is worth more than any amount of correction afterward, and the phone makes that consistency possible for free.

What the AI adds

The phone gives us a good map. The learned model, built by Somnus, does three things with it that the phone alone cannot.

First, it makes the map precise and repeatable. A landmark placed slightly differently on two photos is noise, and noise looks like change. The model is trained to place the same points on the same anatomy across different light, different angles, and different weeks, so that when a point moves, it moved because your face did.

Second, it aligns. It takes today's map and this week's photo and sets them onto your earlier ones so that the small differences that remain after gating, a degree of tilt, a centimeter of distance, a shift in shadow, are corrected out. What is left is you, on top of you.

Third, it reads the soft tissue. Bone does not move over a wellness journey. Soft tissue does. The model follows change through the cheeks, along the jawline, under the chin, around the eyes, and down the contour of the neck. These are the places a weight, GLP, or longevity journey shows first, and often the places people notice in the mirror before the scale catches up.

The model is deterministic. Same photo, same result, every time. It runs on the device. It never paints anything back onto your face. And it does not read your mood, your age, your identity, or anything else it was not built to read. We wrote more about that in How We Use AI.

Why your baseline is the whole point

Every other face product we have seen compares you to a population: an average, a norm, a "score" that quietly means "how you rank against strangers." We think that is the wrong question, and often a harmful one. Your face is not a stranger's face. Your bone structure, your fat distribution, your age, your history, none of it belongs on someone else's curve.

So VisageAir starts by learning you. Your first captures become your baseline: your face, at rest, measured as it is. From then on, every read is against that baseline and against the captures between. There is no leaderboard, no average, no percentile. There is your own line, and there is where you are on it today.

Why over time, and not once

A single photo is a moment, and a moment can lie. Water, sleep, salt, a long day, a flat light: any of them can move a face for an afternoon. A series over weeks cannot be fooled the same way. Trend is honest in a way a snapshot is not, which is why we ask for one capture a week and why we say nothing about your face until there is enough of a series to say something true.

Longitudinal is the technical word for this. It means the same measurement, on the same person, repeated over time, taken the same way each time. It is the method behind most serious long-term health and wellness science, and it is the method here. Week over week, month over month, you against your own earlier self, with every frame gated at capture and aligned by the model so the comparison holds.

What you see is the trend and a plain sentence about it. Steady. Moving. Worth a look. Words, on a card, on your own line. Never a percentage, never a score, never a verdict. We explained why a trend and not a score, and we hold to it.

Where it sits beside your nights

Your face is the day. It is what you and everyone around you can see. Your nights are heard, on the device, by Nightsong. Together they make one honest picture: a face that is changing, and a night that may or may not have changed with it. The face never claims to measure the night. It sits beside it, so you can see both at once.

What it never does

• It never identifies you, or matches your face against any database, anywhere.
• It never leaves the device to be read. The map is built and compared on your phone.
• It never paints colors, labels, or heat onto your face. Results are words on a card.
• It never scores you against anyone else. Your baseline is you.
• It never diagnoses, screens for, or predicts any disease. It is wellness and longevity information only.

A note on your face →

Your face photo is the most identifying data there is, and we treat it that way. What We Do With Your Face says exactly what happens to it. Our Privacy Policy and Terms are the governing documents.

Facial recognition is a serious technology. We think it deserves a serious, honest use. This is ours.

This is part of VisageAir Explained, a series describing who we are, what we make, and the principles that guide our work. All six pieces →

We are glad to talk further about how we think and how we build. Reach us on the contact page.

What Facial Recognition Means Here

---