Evaluation Criteria

What we measure, and how

We measure geometric proportions readable from a single photo and score how close they are to the aesthetic baseline used for the selected gender. Every measured item and its point value is disclosed here in advance. Any feature that can't be reliably measured due to photo quality is marked inconclusive instead of having a number invented for it.

What exactly does the score mean

❌ "78 points = 78 points worth of attractive"

✅ "78 points = the measured proportions are this close to the reference baseline"

This is not physiognomy

Physiognomy (관상) is a traditional East Asian practice of reading personality or fortune from facial features. This site's core result is not physiognomy — it's the geometric proportions actually measured from your photo (eye size, jawline angle, and so on) compared against gender-specific baseline ratios, and every item and point value is disclosed below. If you're curious about a physiognomy-style reading, we've built a separate page that applies your already-measured values to traditional East Asian face-reading lore, purely for fun.

See the physiognomy reading

How the measurement works

Your photo is never sent to a server. Inside your browser, an open-source library from Google (MediaPipe Face Landmarker) locates 468 points on your face, then fixed formulas compute distances and angles between them. This computation is deterministic — uploading the same photo again always produces the same score.

Where do the target values come from

The "ideal" value for each metric comes from two kinds of sources. (1) Classical canons — eye spacing equal to one eye's width, a height:width ratio of 1.3:1, and so on, conventions widely cited in classical portraiture and modern cosmetic-surgery literature. (2) Anthropometric data — eye width, nose width, and jaw (gonial) angle use values from facial anthropometry literature (Farkas et al.; nose width specifically from Farkas, Phillips & Katic, 1998, a Canadian adult average). Neither source uses the "golden ratio" (1.618) — its correlation with attractiveness has repeatedly failed to replicate in aesthetics research, so we don't rely on it. **To be upfront: the anthropometric data is North American/Canadian population statistics, not Korean-specific data.** A few metrics (brow-to-eye gap, iris size ratio, nose bridge height, nose length) don't have a clear conventional or academic source, so their target is instead the average measured across several different faces. Either way, different people may reasonably prefer different proportions; these numbers aren't a claim about what's "correct." Facial-thirds balance (upper/mid/lower) is shown for reference only and isn't scored, since the hairline can't be measured.

Proportion

20pts

Facial Thirds Balance: Measures what percentage each of the face's three vertical sections (hairline-to-brow, brow-to-nose-tip, nose-tip-to-chin) takes up. The hairline isn't something the landmark-detection library can recognize (hair itself isn't part of the face coordinates), so the top section is shown for reference only — scoring is based on the reliably-measurable brow-to-nose-tip vs. nose-tip-to-chin balance instead.
Face Aspect Ratio: Measures how many times the face's height is relative to its width. There's no better or worse value — it's simply a numeric expression of face shape.

Symmetry

20pts

Eye Height Difference: Measures how much the left and right eyes differ in height, as a ratio of face width. Closer to 0% means more level.
Mouth Corner Level: Measures how many degrees the line connecting the left and right mouth corners tilts from horizontal. This can easily shift with expression.

Eyes

15pts

Eye Width: One eye's width, measured in actual millimeters. The iris diameter in the photo (nearly constant across people, about 11.7mm) is used as a ruler to convert coordinates to mm — this lets the value be compared directly against averages from anthropometric literature.
Eye Spacing: Measures how many times the distance between the two eyes is relative to the width of one eye.
Iris Size Ratio: The ratio of the iris (colored part of the eye) diameter to the eye's width — a rough proxy for how large the eyes look relative to the iris.

Eyebrows

10pts

Brow-to-Eye Gap: The vertical distance between the eyebrow and the eye, as a ratio of the face's height.
Eyebrow Symmetry: How much the brow-to-eye gap differs between the left and right sides. Closer to 0% means more symmetric.

Nose

12pts

Nose Width: Nostril width, measured in actual millimeters (iris-based mm conversion, see the eye width description). The target value comes from anthropometric literature (Farkas et al., 1998) — a Canadian adult average.
Nose Bridge Height: How far the nose bridge projects toward the camera relative to the cheek plane, as a ratio of face width. Unlike other metrics, this uses coordinate depth (z), which MediaPipe estimates from a single photo's inferred 3D face shape rather than measuring true depth — treat it as a rough reference only.
Nose Length: The length from the top of the nose bridge to the tip, as a ratio of face height.

Mouth

12pts

Mouth Width Ratio: The ratio of closed-mouth width to the width of the face.
Lip Ratio: The ratio of upper lip thickness to lower lip thickness.

Contour

11pts

Jaw Angle: The angle formed by the jawline running from below the ear to the tip of the chin. A smaller angle means a more angular jawline.
Lower Jaw Width: The ratio of the jaw's left-right width to the width across the cheekbones.

When does a result get marked inconclusive

If any of the conditions below apply, the entire result (all 7 items) is usually marked inconclusive, its point value excluded from the total, and the score rescaled to 100 using the remaining items. Two exceptions: symmetry is marked inconclusive on its own if the face is turned just 5° or more (other categories allow up to 15°) — even a slight turn breaks left-right comparison. Mouth is marked inconclusive on its own if you're smiling or your mouth is open (EXPRESSION) — expression changes mouth shape far more than any other feature, so scoring it would measure the moment's expression rather than the mouth's actual shape.

BLUR: The photo was out of focus, making it hard to accurately measure this feature's lines.
ANGLE: The face was turned too far from front-facing, making it hard to measure this feature's proportions.
LIGHTING: Shadows or overexposure made it hard to distinguish this feature's outline.
RESOLUTION: The resolution around this feature was too low to measure.
EXPRESSION: The mouth measurement was held back because you're smiling or your mouth is open.

Note: obstructions like glasses, a mask, or a hand covering part of the face are not separately detected. The landmark-detection library will still estimate plausible coordinates for a covered feature from surrounding context, so an obstructed photo can produce an inaccurate value without any warning — please follow the photo tips below and avoid covering your features in the first place.

Photo tips

Face the camera directly. A tilt of 15° or more will throw off proportion measurements (the symmetry category is marked inconclusive at just 5° or more).
Shoot in even lighting with no shadows. Backlighting or strong side-lighting makes contours hard to read.
Don't let hair, glasses, a mask, or your hand cover your eyes, nose, mouth, or jawline (as noted above, this case isn't automatically filtered out).
Use a sharp, in-focus photo. Blurry photos are marked inconclusive.
Make sure your face takes up a decent portion of the frame. Too small, and the result is marked inconclusive.
A neutral expression is recommended. A big smile or open mouth will mark the mouth category inconclusive.

Limits of these measurements

This result quantifies the geometric features of a single photo, and can vary significantly depending on the shooting angle and lighting.

This result is not a medical or cosmetic diagnosis — it's an approximation estimated from the 2D information in a single photo. We don't provide rankings, percentiles, or comparisons with other people.