Facial attractiveness test with the working shown

Overall symmetry. Every paired landmark is reflected across the facial midline and the mismatch is measured as a percentage of face width. Feature pairs are then combined - jaw heaviest, nose lightest - into one 0-100 figure. Deviation is mapped to score through a curve calibrated on real faces, so a typical face lands near 72 rather than being punished for ordinary asymmetry.

Golden ratio. Visible face height divided by face width, compared to the classical 1.618 target adjusted for the fact that a landmark mesh cannot see the hairline.

Facial thirds. The three vertical sections should be close to equal. The lower third is the most variable in the population and the one most often flagged.

Facial fifths. Face width should divide into roughly five eye-widths. This is where eye spacing shows up.

Eye and jaw symmetry. Reported separately as well as inside the overall figure, because they are the two pairs the human eye actually notices.

• A single 2D photograph cannot see depth. Projection, chin position and cheekbone prominence are largely invisible to it.
• The proportion targets are inherited from a European-descended aesthetic tradition. They are not neutral, and a lower proportion score can simply mean your face is a different shape than that tradition assumed.
• Skin, hair, grooming and expression are unmeasured, and in real judgements they matter more than any ratio here.
• The percentile bands come from a distribution of self-selected scans - people who chose to test their face - not a random sample of humanity.

The method page carries the full detail and the weights.

What does a facial attractiveness analyzer measure?

It measures geometry: the positions of facial landmarks and the distances, ratios and angles between them. Specifically, bilateral symmetry per feature pair, the height-to-width golden ratio, the balance of the three vertical facial thirds, and the five horizontal fifths. It does not measure skin, hair, expression, styling or anything else that varies between two photos of the same person.

How is the composite score calculated?

Overall symmetry contributes the largest weight, followed by the golden ratio, the thirds balance and the fifths balance. Each is a 0-100 sub-score computed from measured deviation. The weighted composite is then divided by ten and displayed out of 10. Every sub-score is shown next to the composite, so nothing is hidden.

Why is symmetry weighted most heavily?

Because it is the measure with the strongest published link to human attractiveness ratings, and because it is the most robust to landmark noise across different faces. Proportion targets are more culturally loaded, so they carry less weight.

What is a normal symmetry score?

Most faces score between 65 and 80 out of 100 on overall symmetry. Perfect bilateral symmetry does not exist in human faces - everyone is slightly asymmetric, and small asymmetries are not detectable to the naked eye. Scores below about 45 usually indicate a turned or tilted photograph rather than an unusual face.

Can I compare two photos of myself?

Yes, and it is a more informative comparison than a single score. Run one photo, note the numbers, then run another. The gap between the two tells you how much the score depends on lighting, angle and expression rather than on your face - usually a lot more than people expect.

Does the facial attractiveness test store anything?

Nothing. There is no database, no account system and no analytics on your result. The landmark model is fetched to your browser once and then everything, including the photo, stays on your device.