Face Symmetry Test: Free Mirror Analysis of Your Features

What your result shows

This free face symmetry test does one thing: it measures how closely the left half of your face, reflected across a centre line, lands on the right half. A face-landmark model places 478 points on your photo inside your browser, and the tool turns the left–right differences into the numbers on the result card. The same idea is sometimes searched for as a symmetrical face test; either way, the output is a measurement of one photo, not a judgement of a person.

The card also reports three single-purpose readings: how straight the fitted midline is, how far the tip of the nose sits from that line, and how much the line through the mouth corners tilts. A confidence value with one decimal and plain notes on head angle and light tell you how far to trust this particular photo. If the head is turned more than 10° or pitched too far up or down, the card asks for a retake and withholds the scores, because pose creates more apparent asymmetry than most faces contain.

How landmark mirroring works

The method has four steps, and each one is deterministic. First, the eye line is levelled (head roll is removed) and small head turns within ±10° are undone using the model’s depth estimate. Second, a straight midline is fitted through 28 landmarks that run down the centre of the face. Third, each of 220 left/right landmark pairs is reflected across that line. Fourth, the distance between every reflected point and its partner is averaged and divided by the distance between the pupils (iris centres, landmarks 468 and 473), so the size of the photo does not matter.

The 220 pairs are not hand-picked. They are derived from MediaPipe’s canonical face model, which is an exact mirror image of itself, so every landmark on one side has a defined twin on the other. On that canonical face the tool reports 0% mirror error, which is the check that the pairing and the reflection are correct. Real faces always sit above zero.

These five regions hold 68 of the 220 pairs. The remaining pairs, from other parts of the face mesh such as the cheeks and forehead, still count towards the overall mean mirror error but belong to none of the five region rows. The three extra readings are separate measurements. Midline straightness shows how much the 28 midline points wander from the fitted straight line. Nose-tip offset is the sideways distance of the nose tip from that line. Mouth-line tilt is the angle of the line between the mouth corners 61 and 291 relative to the level eye line.

How the symmetry score is built

The score starts at 100 and loses 6 points for every 1% of pupil distance in mean mirror error, with the result kept between 0 and 100. As arithmetic examples, an error of 1.0% gives 94, 2.0% gives 88 and 3.0% gives 82. The constant of 6 is ours: no published norm exists for a landmark-based mirror error, so it is chosen to spread everyday photos across a readable range rather than taken from a study.

Why we do not show “left–left” and “right–right” faces. Many symmetry apps build two composite portraits from one half of the face mirrored onto itself. We deliberately do not. The tool generates no images; it only measures. Composites are easy to misread as “what you would look like”, and they depend heavily on exactly where the midline is placed. Numbers per region tell you more and cannot be mistaken for a different face.

How to read your symmetry result

When you run a face symmetry test on your own photo, start with the regional table rather than the total. The “most matched” and “least matched” labels are relative to your own photo: one region has to come first and one last, even when all five are close. Compare the actual percentages, and the gap between them, before reading anything into the order.

Some asymmetry is normal in everyone. Research on symmetry and perceived attractiveness, such as Perrett et al. (1999) and Rhodes et al. (1998), is built on ordinary faces that all differ slightly between sides, and lay medical overviews describe mild asymmetry as common. Symmetry is also only one ingredient of how faces are perceived (Jones & Jaeger 2019), which is why the overall score gives it 35% rather than all of it.

Confidence matters as much as the number. A higher mirror error with a low confidence value usually points to the photo: a turned head, light from one side or an uneven expression. A similar value that stays similar across several well-taken photos reflects the face geometry. Expression is a common cause of a gap between photos: a half-smile lifts one corner of the mouth and tilts the mouth line, even though the resting face has not changed.

A calm note on health. This tool describes a photo and cannot diagnose anything. If you notice a sudden change on one side of your face, such as drooping, weakness or numbness, contact a doctor or emergency services promptly rather than relying on a photo measurement.

Photo conditions that create false asymmetry

Any face symmetry test reads two things at once: the face and the photo. These habits reduce the second.

Symmetry is one of four parts of the overall face rating. To see how the other parts are measured, open the golden ratio calculator, the face shape detector or the jawline analysis, which looks at the jaw width and angle from the front.