Golden Ratio Face Check

One photo, five facial proportions. For each one you get the measured value, the target this page uses, the tolerance band, the deviation from the target, and how close that is on a 0–100 scale. The model runs in your browser and the photo is never uploaded.

Drop a front-facing photo here, or use Choose a photo. JPG, PNG or WebP. One face, hair off the forehead, even light.

The five ratios, and what this page targets

Every number below is this page's own heuristic, chosen and published by us. None of it is a published clinical, medical or official standard, and nothing here is copied from another tool.

RatioMeasured betweenTargetToleranceWhere the target comes from
Face length ÷ cheekbone widthlandmark 10 → 152, over 234 → 4541.618± 0.30The golden ratio itself (φ)
Forehead width ÷ cheekbone widthlandmark 127 → 356, over 234 → 4540.88± 0.12The mean of the seven published profiles in face-shape-ratios.csv
Jaw width ÷ cheekbone widthlandmark 172 → 397, over 234 → 4540.84± 0.12The mean of the seven published profiles in face-shape-ratios.csv
Mouth width ÷ nose widthlandmark 61 → 291, over 129 → 3581.618± 0.25The golden ratio itself (φ); the wider band is our own choice
Eye spacing ÷ eye widthlandmark 133 → 362, over 33 → 1331.00± 0.20The classical "eyes one eye-width apart" spacing

All distances are measured in the pixels of your own photo (scaled by its width and height), so the ratios do not depend on how large the photo is. Landmark numbers refer to the same 468-point face landmark mesh the face shape detector uses; the index list is in golden-ratio-landmarks.csv.

How the deviation and the proximity number are worked out

  1. value = segment A ÷ segment B, where A and B are the two landmark pairs in the table above.
  2. deviation = value − target.
  3. inside the band when the absolute deviation is less than or equal to the tolerance.
  4. proximity = 100 × (1 − |deviation| ÷ (2 × tolerance)), clamped to 0–100. It reads 100 on the target and 0 once the ratio is two tolerances away.
  5. mean proximity = the mean of the five proximity values, rounded to a whole number.

The proximity number is a closeness measure against this page's own targets. It is not a rating, not a probability and not a percentage of anything.

Printed from faceshapehub.com/golden-ratio-face/ — targets and tolerances are this page's own heuristic values.

Take the numbers with you

What this page is not

FAQ

Is this a golden ratio face test that scores beauty?

No. It reports five ratios and how far each sits from the value this page targets. The numbers are not a rating, not a probability and not a verdict on attractiveness.

Does my photo get uploaded?

No. The landmark model is a local file served from this site and it runs in your browser; the photo never leaves the device.

Where do the target values and tolerances come from?

They are this page's own heuristic values. The 1.618 targets come from the golden ratio itself, 0.88 and 0.84 are the mean of the seven profile values published in the data files, and 1.00 is the classical one-eye-width spacing. The tolerances 0.30 and 0.12 match the published tolerances, and 0.25 and 0.20 are this page's own wider bands. None of them is a published standard.

Why do my numbers change between photos?

The landmark points move with lighting, head angle, expression, glasses and lens distortion, so the ratios move too. A straight-on photo with even light gives the most stable result.

Can I check this by hand instead?

Yes. The tape-measure walkthrough covers measuring a face by hand and includes a sheet you can print and fill in.

Built and maintained by one person under the site name Face Shape Hub. The About page explains what this site is and what it deliberately does not do; the contact page is where wrong results and broken pages get reported.