Midface Ratio: Measure It and Know What the Number Means

The scan finds the centre of each pupil (MediaPipe's iris landmarks) and the top edge of the upper lip on the midline. It levels the image so the pupils sit on a horizontal line and turns the head back to frontal. Then it divides two distances: the straight-line distance between the pupils, and the vertical height from the pupil line down to the top of the upper lip.

Midface ratio = pupil distance ÷ pupils-to-upper-lip height. A value of 1.00 means the two distances are equal. A value of 1.10 means your pupils are 10% farther apart than the height below them; 0.90 means the height is the larger of the two.

We use this direction because it's the one used in the forum threads where most people first meet the term, so your number can be compared with what you've read there. Higher numbers mean a wider, shorter midface; lower numbers mean a taller, narrower one. Switch on the Ratios layer to see the pupil line and a vertical line running down from it toward your mouth; the height used in the ratio stops at the top edge of your upper lip.

Search results don't agree on what the midface ratio is, which is why two sites can give you numbers that look contradictory. Here's how they differ and whether you can convert.

Not one that's been published. We looked for peer-reviewed studies that measure pupil distance ÷ pupils-to-upper-lip height in a population, or test it against attractiveness ratings, and found none. Forum threads describe an “ideal” near 1.0, with titles like “ideal midface ratio is 0.95–1”, but we couldn't trace that figure to a study. A 2026 commentary in Plastic and Reconstructive Surgery Global Open by Maini and Janis lists the formulaic ratios circulating in looksmaxxing culture and warns that they're often understood partially and misapplied.

Half of the ratio is well documented. Adult pupil distance has been measured many times; for example, it averaged 68.09 mm in men and 63.6 mm in women in a photographic study of 300 Malaysian South Indian adults (Packiriswamy and colleagues, 2012). But we found no matching large-sample norm for the pupils-to-upper-lip height, so a population ratio can't be built from published data.

A close cousin shows why definitions matter so much. The facial width-to-height ratio (fWHR) is also width divided by a mid-face height. With manually placed landmarks, Carré and McCormick (2008) found averages of 1.80 in women and 1.86 in men; with automatic landmarks on 17,607 passport photos, Summersby and colleagues (2022) found 2.18. Same idea, different points, a gap of more than 0.3. A meta-analysis also found the difference between men and women in fWHR is very small (d = .11; Geniole and colleagues, 2015). Any midface “ideal” quoted without its landmarks and sample deserves the same caution.

So FaceCheckup shows it as a plain number with a ± range. We don't call it good, bad, compact or long, because no source we trust supports those labels.

The Ratios layer draws the lines this page is about. The yellow line runs between the pupils. The teal line drops from the pupil line to the mouth. The dashed red line is face width. We tested it on a public-domain painting, Leonardo's Mona Lisa, rather than on a real person.

On that image the scan measured a midface ratio of 0.90 ± 0.02. Pupil distance came out at 43.4% of face width, and the eyes-to-mouth distance at 39.8% of face length. Because the photo check also found a 17° head turn and a 19° chin tilt, the scan treated the photo as a retake. A turned head shortens the pupil line on camera, which lowers this ratio. Re-scan straight on before comparing your own ratio with anything.

Because the midface ratio divides a horizontal distance by a vertical one, anything that stretches the face vertically in a photo changes it. Four things matter most.

Chin tilt: tipping your head up or down foreshortens the height in a flat photo. The quality check warns beyond 10° and asks for a retake beyond 18°. Camera distance: close selfies distort the centre of the face; Pressler and colleagues (2022) measured nasal length 6.4% longer in 12-inch selfies than in clinical photos taken from 5 feet, and the nose sits right inside this measurement. Expression: a smile moves the upper lip, which is one end of the height. Resolution: below about 60 pixels between the pupils, landmark placement gets coarse, so the scan warns you.

On top of that, the landmark model's own average error is about 2% to 3.3% of the distance between the eyes, per Google's model card. That's why the scan reports a ± range from four framings and never shows the ratio to more than two decimals. If two of your photos differ by less than their ranges, they're the same result.

Here's how to read the result card, using made-up numbers rather than anyone's real face. Say the scan shows Midface ratio 1.06 ± 0.02. That means across your photo and its three re-framed copies, the ratio ran from about 1.04 to 1.08: your pupils are a little farther apart than the height from the pupil line to the top of your upper lip.

Now say a second photo, taken the next day in the same spot, shows 1.03 ± 0.02. The ranges (1.04–1.08 and 1.01–1.05) overlap, so the honest conclusion is that both photos measured the same thing. If a third photo came back at 0.97 ± 0.02, check its quality notes before believing it; a tucked chin or a slight smile is the usual reason one photo stands apart.

It also helps to read it next to the two ratios it shares distances with. Pupil distance ÷ face width uses the same top line, and eyes-to-mouth ÷ face length uses almost the same height, measured down to the mouth line instead of the top of the lip. If your pupil-distance share is on the high side and your eyes-to-mouth share on the low side, this ratio will come out high, and the reverse. Those two are the ratios with a published reference (Pallett, Link and Lee, 2010), which is why the Balance Score uses them and not this one.

Use one good photo rather than many quick ones. These steps follow our rulebook's measurement rules.

Honestly, not much, and that's a relief. Without a population norm or an attractiveness study behind it, the ratio is a description of one photo, not a score. The two proportions that were tested on raters, pupil distance ≈ 46% of face width and eyes-to-mouth ≈ 36% of face length (Pallett, Link and Lee, 2010), are already in your Balance Score, and they overlap with what the midface ratio tries to capture.

You'll see claims online that mewing, “bonesmashing” or exercises can shorten the midface. A 2025 systematic review found the evidence for mewing sparse and mostly anecdotal (Seraj and colleagues), and a 2026 study flagged bonesmashing as a dangerous trend spreading on social media (Grillo and colleagues). This site doesn't give medical advice; if you're concerned about your facial structure, talk to a licensed dentist or doctor.

What does change how your midface looks in photos is the camera. Our rulebook's research-supported tip is to shoot from about 1.5 m with the zoom instead of at arm's length; in experiments, the same faces photographed at 45 cm were rated less attractive, trustworthy and competent than at 135 cm (Bryan, Perona and Adolphs, 2012). Photographers also suggest holding the camera at eye level or slightly above, never below the chin.

If you're exploring the eye area, our canthal tilt page measures eye angle the same careful way, high cheekbones covers where the widest point of your face sits, and the PSL scale page explains the ranking culture these ratios come from.