AI face analyzer. See what stands out about your face.

The face above, in numbers

These update as you move the dots or switch faces.

• Face width (FWHR): cheek-to-cheek width ÷ eyelid-to-lip height1.92 · men average 1.86, usual 1.73–1.99
• Eye tilt: how high the outer eye corners sit above the inner onesAbout +1° (estimate, ±1.5°) · most men +1° to +5°; up to 1.5° past either end is within the reading’s margin
• Face thirds: hairline to brows, brows to nose base, nose base to chinForehead 26% · middle 36% · lower 38%

The points behind the three readings

CEVO places a 478-point face mesh (MediaPipe numbering) on the photo, plus a named hairline point that the mesh lacks. On your own photo, both are found in your browser: MediaPipe Face Landmarker finds the mesh, and the hairline is where MediaPipe's hair-and-skin segmentation model first sees hair above the top of the mesh, along the centre of the forehead. The photo is never uploaded; only the two models are downloaded. If no hairline is found (hair pulled far back, a hat, a shaved head), the thirds wait for you to drag the top dot. The example faces were read by the same pass, so an example and the same photo uploaded give the same result.

Face width (facial width-to-height ratio, FWHR)

The widest points of the face outline near the cheekbones (#234 to #454), a stand-in for bizygomatic width, divided by upper face height, from the highest point of the upper eyelids (halfway between #159 and #386) to the top of the upper lip (#0). Forehead and chin are left out. Weston et al. (2007) describe the measure on skulls; photo studies take the height from the upper lip to the brow line or to the top of the eyelids, and Carré and McCormick (2008) measured it on photos of 88 students, the averages we compare with. Kramer (2017) pooled the published studies and found the difference between men and women small.

Eye tilt (canthal tilt)

The angle of the line from the inner eye corner to the outer corner (#133 to #33 on the right eye, #362 to #263 on the left), measured against the line through both pupils (#468 to #473). Using the pupil line cancels a tilted head. Positive means the outer corner sits higher. CEVO's reports give eye tilt as an estimate, about ±1.5° (its terms), so a 1° difference between the eyes is within its margin. The reading shown is the two eyes' average, in whole degrees.

Face thirds (vertical facial thirds)

Hairline (trichion) to the brow centre (#9), brow centre to the base of the nose (subnasale, #2), and base of the nose to the chin (menton, #152), each measured along the hairline-to-chin line and shown as a share of the whole. Farkas et al. (1985) found that real faces rarely split into three equal parts, so the reading compares the three parts of the same face with each other: it names the part furthest from a third when it is 4 or more points of the whole away, and says "close to even" otherwise.

The averages we compare with

"About average" means within 0.4 SD of the mean; "a little wider than average" 0.4 SD up to 0.03 past the typical range; "wider than most" more than that. Face width is compared at two decimals, as in the table. Eye tilt words follow whole degrees: +5° and up tilts up, +2° to +4° tilts up a little, 0° to +1° is about level, below 0° tilts down; once men or women is known, the eye tilt is also checked against the typical range below, with 1.5° of leeway, the estimate's margin. "What stands out" names a reading more than that leeway outside its typical range, or a third 4 or more points from even, furthest first; when none is, it names, in softer words, the reading whose card puts it most strongly: a width a little wider or narrower than average, or an eye tilt just past its typical range but within the leeway, comes before an eye tilt in the outer quarter of its typical range or a third 3 or more points from even; with none of these, it says nothing stands out. The leeway keeps a dot a pixel off, or a second selfie, from changing the answer. A narrower width from your own photo is not counted, because an arm's-length selfie makes faces read narrower.

Typical range is the mean ± 1 SD (the 16th to 84th percentile), so about 1 in 3 adults sits outside it. Face width averages from Carré and McCormick (2008), photos of 37 men and 51 women; eye tilt averages from Farkas et al. (2005) and Rhee et al. (2012). The thirds have no average here: they compare the parts of the same face.

What the camera does, and the iris check

A phone's wide front camera at arm's length (about 30 cm) is far closer to the nose than to the jaw. By our camera model (see why selfies change your face), the nose reads about 4% wider at 30 cm (5% at 25 cm) and the jaw about 3% narrower than from 1.5 m. Ward et al. (2018) calculated the nose base about 30% wider relative to the cheekbone width at 12 inches than at 5 feet; Třebický et al. (2016) showed that focal length changes how a face's shape is seen. Adult irises (the white-to-white width of the cornea) are about 11.7 mm ± 0.4 mm wide (Rüfer et al. 2005), so their size in pixels shows how far away the camera was. When the photo is extremely close (under about 25 cm), CEVO asks for a retake; a normal selfie is accepted, and the scan marks the values a close photo can shift. It never warps or edits the face.

What is measured, and what is not

Measured: distances and angles between landmark points, in source-image pixels, turned into ratios so the photo's size does not matter. Compared: face width and eye tilt against adult averages for men or women, picked by you on your own photo; the thirds against each other. Not done in the free scan: no score and no ranking against other people; the paid reports add a facial balance breakdown, feature by feature, with no single overall rating. No edits to the photo. Dragging a dot on this page changes only that reading, live, in your browser.

How the 169 tests split by region

Sources

1. Weston EM, Friday AE, Liò P (2007). Biometric Evidence that Sexual Selection Has Shaped the Hominin Face. PLoS ONE, 2(8), e710.

   DOI: 10.1371/journal.pone.0000710

2. Carré JM, McCormick CM (2008). In your face: facial metrics predict aggressive behaviour in the laboratory and in varsity and professional hockey players. Proceedings of the Royal Society B: Biological Sciences, 275(1651), 2651–2656.

   DOI: 10.1098/rspb.2008.0873

3. Kramer RSS (2017). Sexual dimorphism of facial width-to-height ratio in human skulls and faces: A meta-analytical approach. Evolution and Human Behavior, 38(3), 414–420.

   DOI: 10.1016/j.evolhumbehav.2016.12.002

4. Farkas LG, Katic MJ, Forrest CR, et al. (2005). International anthropometric study of facial morphology in various ethnic groups/races. Journal of Craniofacial Surgery, 16(4), 615–646.

   DOI: 10.1097/01.scs.0000171847.58031.9ePMID: 16096529

5. Rhee SC, Woo KS, Kwon B (2012). Biometric study of eyelid shape and dimensions of different races with references to beauty. Aesthetic Plastic Surgery, 36(5), 1236–1245.

   DOI: 10.1007/s00266-012-9937-7PMID: 22806145

6. Farkas LG, Hreczko TA, Kolar JC, Munro IR (1985). Vertical and horizontal proportions of the face in young adult North American Caucasians: revision of neoclassical canons. Plastic and Reconstructive Surgery, 75(3), 328–338.

   DOI: 10.1097/00006534-198503000-00005

7. Ward B, Ward M, Fried O, Paskhover B (2018). Nasal Distortion in Short-Distance Photographs: The Selfie Effect. JAMA Facial Plastic Surgery, 20(4), 333–335.

   DOI: 10.1001/jamafacial.2018.0009

8. Třebický V, Fialová J, Kleisner K, Havlíček J (2016). Focal Length Affects Depicted Shape and Perception of Facial Images. PLoS ONE, 11(2), e0149313.

   DOI: 10.1371/journal.pone.0149313

9. Rüfer F, Schröder A, Erb C (2005). White-to-white corneal diameter: normal values in healthy humans obtained with the Orbscan II topography system. Cornea, 24(3), 259–261.

   DOI: 10.1097/01.ico.0000148312.01805.53