Eye separation ratio: eyes versus face width

What eye separation ratio is

Eye separation ratio is the distance between your pupils as a share of the width of your face. It captures how much of the face width the eyes span. In the app this measure is labelled Eye-to-face width.

A higher value means the eyes take up more of the face width and the cheeks and temples feel narrower. A lower value means a broader face around the eyes. Like most proportions, it describes a style of face, and the range LookFlex uses marks a balanced middle of its reference set.

In the LookFlex app this measurement appears as Eye-to-face width.

How LookFlex measures it

LookFlex measures the distance between the centres of your two pupils, using the iris landmarks, and divides it by the width of the face at eye level, taken between the outermost points of the cheek contour on each side.

Because both distances are measured in the same photo, the ratio does not depend on image size. It does depend on a level, front-facing capture: if your head is turned, one side of the face is foreshortened, which is why LookFlex corrects for pose and lowers its confidence when the angle is large.

The calculation: ratio = distance between pupils ÷ face width at eye level.

Landmarks used

LookFlex reads these points from the 478-landmark face mesh. The numbers are each point's index in that mesh.

• Pupil centres: 468, 473
• Widest points of the face at eye level: 234, 454

The full pipeline, including how pose and mirror-averaging reduce noise, is on how LookFlex measures your face.

Typical and scoring ranges

LookFlex shows two ranges. The typical range is the middle half of its reference set, from the 25th to the 75th percentile: half of the faces fell inside it. The scoring range is where a reading earns a closeness score of 100; outside it the score falls smoothly, with no cliff at the edge. The scoring range is the same for everyone, while the typical range follows each group's own median. Unit: unitless ratio.

How to read your result

"The range" here means the scoring range. The impressions below are common styling descriptions, not research findings. What research does show is in the next section.

• Below the range: A broad face around the eyes, with a lot of cheek and temple width. It reads strong and grounded.
• Inside the range: Eyes and face width sit inside the scoring range.
• Above the range: The eyes span more of the face width. It reads youthful and balanced, sometimes delicate.

What the research says

Limited evidence A single experiment points to a typical value, close to an average face.

• Faces looked best when the gap between the eyes was about 46% of face width, close to an average face (Pallett et al., 2010). Studies use different landmarks from LookFlex's.
• Face width relative to height tracks body mass index, so soft tissue can move width-based measures (Coetzee et al., 2010).

No study we cite tests LookFlex's own ranges. They come from the calibration described on the methodology page, and every source is listed under References below.

What influences it

Mostly fixed

• The width of the cheekbone arch and the spacing of the eye sockets, which are largely inherited.
• Age, which can change the soft tissue at the edges of the face.

Things that move the reading

• Cheek fullness. The face-width landmarks sit on the skin contour, so soft tissue, body composition and water retention move them a little.
• Camera distance. A very close selfie enlarges the centre of the face relative to its edges.
• Head pose and expression, which shift the contour points.

How to improve how it reads

These suggestions change how the measurement reads or how well the photo captures it. Styling tips are common practice rather than tested in research, and none of them promises a change in bone structure.

• Hold the camera at arm's length or further, with the lens at eye level, so the lens does not distort the centre of the face.
• Across people, face width tracks body mass index (Coetzee et al., 2010), so body composition can move this reading a little. If you want to change it, a steady routine of strength training and a balanced diet is the route we suggest, not a crash diet.
• Hairstyles that add or remove width at the temples change how the proportion reads without changing the number.

Common myths

Myth: There is one scientifically perfect value, 0.46.

There is no single scientifically perfect value. One experiment found faces looked best when the gap between the eyes was about 46% of face width, close to an average face (Pallett et al., 2010), but studies use different landmarks. LookFlex's face-width points sit slightly wider than true bone width, so its reference median is about 0.45 and the scoring range is 0.435–0.475 for everyone.

Myth: Face width is all bone.

The contour includes skin and soft tissue, so it can shift a little with body composition and camera angle.

How it feeds your score

This measurement is one input to 2 sub-scores, which then combine into your Glow Score:

• Cheekbones: 10% of the sub-score. Cheekbones carries 12% of the Glow Score.
• Harmony: 10% of the sub-score. Harmony carries 22% of the Glow Score.

Precision. Every reading of this measurement comes with a 95% margin, estimated by measuring your photo three times. Across the reference set the median margin was ±0.0079: a retake in similar conditions usually lands within that much of the first reading. A change between two scans counts as real only when it is bigger than both scans' margins combined.

Confidence. LookFlex starts this measurement at a confidence of 90% and lowers it when your head angle, the lighting or image quality weakens the reading. The weight of each input in a sub-score scales with its confidence. Details are on the methodology page.