Free facial analysis from one photo

A face analysis test measures the geometry of your face from a photo and reports it as numbers: the outline and its length against its width, how the face divides into thirds, the angle of each eye, the width of the nose against the gap between the eyes, how closely the two sides mirror each other and how far a set of classical ratios sits from its targets. This one places 478 points on a front photo inside your browser, straightens the head angle first, and then reads every number from those points. It describes proportion. It does not diagnose anything, it is not a medical or cosmetic assessment, and it does not rate how a face looks. Each reading links to the tool on this site that measures that one thing in more depth.

Yes. The whole dashboard is free, with no account and no upload: the face model downloads once to your browser and the measuring happens on your device. You can run it as often as you like, on as many photos as you like, and on an AI-generated example face if you would rather not use your own. Under the dashboard there is an optional paid report. It turns the same numbers into a written plan for the hair and glasses shapes that suit your measured proportions, with hairstyle try-ons, for one payment that does not renew. Nothing on the free dashboard is held back to push you towards it: every reading this photo can support is printed, and the ones it cannot are marked as withheld with the reason.

It is as accurate as the photo allows, and the photo moves the numbers more than most people expect. A head turned a few degrees, a slight smile, side lighting or a phone held close to the face all shift landmark ratios. This page corrects head rotation before measuring and measures each photo five times to find its own noise floor, and it withholds a left-right reading when the difference is smaller than that floor. Face length also includes an estimated allowance for the hairline, because the mesh stops below it. The words beside each number compare it with the middle half of an AI-generated test set, which is a rough reference rather than a population norm. Two photos of the same face taken a minute apart can differ slightly, so compare readings across photos rather than trusting one decimal place.

Because printing a number the photo cannot support would be guessing. Symmetry and canthal tilt are the two readings most easily invented by the photo itself. Each photo is measured five times with tiny shifts, and how far the points move between passes becomes that photo's own measurement floor. If the left-right offset is smaller than the floor, or an eye corner is closer to level than the floor can resolve, the reading is withheld rather than shown. A photo that is slightly sheared, from a wide lens or a screen re-photo, withholds both, because shear manufactures exactly this kind of left-right difference. A sharper, straight-on photo in even light usually brings the withheld readings back. The rest of the dashboard is still measured as normal.

No, and this one does not try. Every reading on the dashboard is a distance or an angle between points on your face, compared with targets or with a reference set, and none of them is a verdict on how you look. Classical targets such as the golden ratio come from drawing conventions, not from research on what people find attractive, and most real faces miss several of them by a wide margin. The research on attractiveness points at averageness, symmetry and sexual dimorphism, and even those explain only part of how people rate faces. If you want a score built from those factors, with its formula and error range published, the attractiveness test on this site does that separately. This page simply shows what your proportions are.

No. The face model downloads once from this site and runs inside your browser tab, so the measurement happens on pixels already in your device's memory. The photo, the 478 points and the readings on the dashboard are never sent to a server and are gone when you close the tab. The page counts anonymous steps, such as that a measurement finished, without any number from your face. If you buy the written report, it is built from measurement numbers only, not from the photo. An optional paid visual has its own explicit consent step before any photo is sent anywhere. If you would rather not use a photo of yourself at all, the example button runs the same analysis on an AI-generated face.