Anthropometry · Multi-Ethnic Framing

East Asian facial features

By Randy · RealSmile

Facial Analysis Research

Verified

Updated

August 8, 2026

How a face with East Asian ancestry scores across 17 structural metrics. Descriptive anthropometry, not a hierarchy.

Beauty is multi-ethnic. Published cross-cultural research finds no single ethnicity scores higher in aggregate. This page describes the metric distribution, not a ranking.

17 metrics · Farkas international norms · Free · No signup

Get My Score →

Free score · $14.99 unlocks population-appropriate percentile report

Why most face tools mis-score East Asian faces

Most legacy face-scoring tools were calibrated on Caucasian-dominant datasets, which means the eye aspect ratio, nasal index, and midface projection thresholds were set against one population's distribution. When a monolid or folded East Asian eye is scored against a Caucasian eye-aperture norm, the resulting percentile reads as a low score on a metric where the face is actually average for its population. Choe et al. (2004) on Korean craniofacial norms and Le et al. (2002) on Vietnamese norms both document distributions that diverge meaningfully from Western reference datasets on a handful of specific metrics.

The fix is not to weight populations differently in the composite. The fix is to show both percentiles. A useful East Asian face report shows you where you sit against the universal distribution (cross-population, descriptive) and against the population-appropriate distribution (Farkas international atlas plus Korean and Japanese datasets where validated norms exist). A metric where you sit at the 35th universal percentile but the 65th population-appropriate percentile is not a weak metric; it is a metric where the universal norm was the wrong reference.

The 17 structural metrics themselves are universal geometry. Facial thirds, fifths, FWHR, canthal tilt, jawline ratio, and the rest are defined the same way regardless of population. What changes is the reference distribution they get compared against. The free composite uses the universal distribution; the $14.99 paid report adds the population-appropriate one.

5 structural patterns documented in East Asian craniofacial research

Bizygomatic to bigonial ratio

Korean and Japanese norms (Choe 2004; Wang et al. 2011) document a tendency toward a wider bizygomatic relative to bigonial measurement compared to European norms, contributing to the perception of higher cheekbones. Wide individual variance.

Midface projection

International atlases (Farkas et al. 2005) record slightly flatter midface projection on average in East Asian populations compared to European samples. Descriptive, not evaluative; flatter and projected both register as neutral on the composite when measured against population-appropriate norms.

Nasal index

Nasal width to nasal height ratio sits in a distinct sub-range. The score treats this as population-appropriate rather than as a deviation from a Caucasian norm.

Eye aperture geometry

Monolid and double-lid eyes register slightly different eye aspect ratios. Population-appropriate norms (Choe 2004) prevent the Caucasian eye-aperture mean from being treated as a universal target.

Lower-face proportion

Average lower-face height relative to total facial height clusters slightly shorter in published East Asian norms vs European samples. Carried as descriptive context in the metric layer.

Hair and brow line

Darker, denser eyebrow hair and a typically lower hairline shift how the brow-to-eye distance metric reads compared to lighter-brow populations. The score treats this as structural signal rather than as a styling artifact.

📊

Recommended next step

Get the Breakdown Behind Your Score

Instead of a single number, see 17 individual metrics — jawline, canthal tilt, symmetry, and more.

Why dual percentiles matter more than a single composite

A single composite forces a face into one reference distribution. For most users that distribution is implicitly Caucasian because the underlying training data was. The result is a percentile that systematically misreads any face whose population norms differ meaningfully from the training set. We avoid this by carrying both percentiles in the paid report.

The universal percentile answers: where do you sit relative to faces of all populations? The population-appropriate percentile answers: where do you sit relative to faces with similar ancestry? Both are descriptive. Neither is a verdict. The gap between them is itself diagnostic; a large gap usually points to a metric where the universal distribution is dominated by one population's mean.

Honest limits of any ethnicity-aware face score

Ancestry is a continuum. Mixed-ancestry faces do not cleanly belong to a single distribution, and the population-appropriate percentile is least useful in that case. The universal percentile remains the more honest read for mixed-ancestry users.
Published norms (Farkas 2005, Choe 2004, Le 2002, Wang 2011) sample a small number of subgroups within each population label. East Asian as a category covers Korean, Japanese, Han Chinese, Vietnamese, and many other populations with distinct craniofacial distributions. Treat the population-appropriate percentile as directional, not as a precise read.
The composite measures still photography. In-person impressions involve motion, voice, and posture none of which a photo captures. A high composite does not guarantee in-person impact and a low composite does not preclude it.
The score does not, and should not, recommend interventions targeting ethnic features. Eyelid surgery, midface implants, and similar procedures carry surgical risk and rarely move the underlying composite by more than a few points relative to their cost. The recommendation layer in the report targets soft-tissue and grooming levers that work across populations.
Beauty is multi-ethnic. The published cross-cultural preference research is unambiguous on this. Any tool that claims one population scores higher than another in aggregate is wrong about the underlying research.

East Asian facial features FAQ

Does this page rank East Asian faces against other ethnicities?+

No. Beauty is multi-ethnic and the published cross-cultural preference research (Cunningham et al. 1995; Rhodes 2006; Coetzee, Greeff, Stephen and Perrett 2014) finds no single ethnicity scores higher than another in aggregate. This page describes how the 17 structural metrics tend to distribute across faces with East Asian ancestry, not which population is more attractive. The composite score and percentile use the population-appropriate distribution as a reference rather than a universal one.

Which structural metrics tend to differ in East Asian faces?+

Anthropometric work (Farkas et al. 2005 on international craniofacial norms, Le et al. 2002 on Vietnamese norms, Choe et al. 2004 on Korean norms) documents directional differences in bizygomatic width-to-bigonial width ratio, midface projection, nasal index, and epicanthic fold prevalence. These are descriptive averages with very wide individual variance. Any individual East Asian face can sit anywhere in the distribution. The 17-metric report shows where a specific face sits, not where a population averages.

Are the percentiles compared to East Asian norms or universal norms?+

Both. The free composite shows your score against the universal cross-population distribution. The $14.99 Looksmax Report adds a per-metric percentile against the published East Asian distribution where validated norms exist (Farkas international atlas, Korean and Japanese craniofacial datasets). The dual percentile is the useful read because cross-population norms can flag a metric as low when the population-appropriate norm would flag it as average.

Does the test handle epicanthic folds and monolids correctly?+

The 68-landmark detector traces the visible eye outline including the medial canthus. Monolid eyes and folded eyes register slightly different eye aspect ratios because the visible aperture geometry differs. The published Korean and Japanese eye-aperture norms (Le 2002, Choe 2004) are coded into the population-appropriate distribution so a monolid eye is not penalized as a low EAR against Caucasian-distribution norms. This is the single most common scoring failure on legacy face tools built only on Western datasets.

Is the test biased toward Western beauty standards?+

The geometry itself is universal. The interpretation is where bias enters, and we mitigate it by showing both universal and population-appropriate percentiles. The expression scoring (Ekman and Friesen FACS) is cross-culturally validated. The recommendation layer in the paid report is structured around levers that work across populations (skin clarity, sleep, posture, grooming) rather than around features that would push a face toward a specific cultural ideal.

Is my photo uploaded?+

On desktop, the 68-landmark detector runs in your browser via a model bundled with the page and the 17-metric vector is computed on your device — your photo is never stored. On mobile, your photo is processed in memory by our scan server and deleted immediately after the landmarks are measured — never written to disk, never stored, and never used to train any model.

What does the free score include and what does the $14.99 report add?+

Free: composite 0 to 100, universal percentile, and your two strongest plus two weakest metrics. Paid ($14.99 Looksmax Report): every metric percentile against both universal and East-Asian-distribution norms, a 5-page written breakdown of which metrics are doing structural heavy lifting in your specific face, and a soft-tissue-first improvement plan.

Free score is the headline. Population-appropriate percentile is the plan.

Get all 17 metrics with dual-percentile context.

The $14.99 Looksmax Report scores all 17 metrics with both universal and population-appropriate percentiles, identifies your two weakest metrics, and writes a soft-tissue-first plan.

Get My Full Report — $14.99 →or try the free score