FACE FORM FINDER · 6 OCTOBER 2026

Face Symmetry Test

A face symmetry test compares matching features on the left and right of a midline, and the number it returns carries two differences at once: the difference the face has, and the difference the camera adds. The camera's share can be written down. A head is a solid about 190 mm front to back, so its two sides sit about 95 mm either side of the axis it turns on. Turning it 2 degrees brings one side 3.3 mm closer to the lens and takes the other 3.3 mm further away. On the fixed face below the two sides are level by construction, so whatever a tool reports is the pose plus the tape — and a 5 mm left-right reading is what a 3 degree turn produces on its own.

So the useful question is not whether a difference is there. It is how much of the difference the camera is entitled to. The tables below put that share in millimetres, on a face whose two sides are equal by construction, so that anything a test reports has to come from somewhere other than the face.

One number, two differences

A symmetry score is built by reflecting each left-hand landmark across a fitted midline and measuring how far it lands from its right-hand partner. That residual is real arithmetic, and it is arithmetic on a photograph. Two things go into it that have nothing to do with the face: where the camera stood when the photo was taken, and how well the landmarks could be found at all. The rest of this page is about the first of those, because that part can be given a figure.

The fixed head, in millimetres
Distancemm
Head, front to back190
Half of that — the lever arm on each side of the turn axis95
Face width, cheekbone to cheekbone118
Face height, hairline midpoint to chin185

These are a defined reference, not a person. A real head will not sit on these numbers, and the point of fixing them is that the arithmetic can then be checked rather than taken on faith.

The four pairs, on a face that is level by construction

A landmark test compares pairs. The four below are the pairs a front-facing photo can see, and on this face each pair is built to the same distance on both sides. That is deliberate: it means the left-minus-right column starts at zero, and stays at zero until something other than the face moves it.

Four pairs, level on the reference face
PairFrom the midline, each sideLeft readingRight readingLeft minus right
Inner eye corner16 mm16 mm16 mm0
Pupil centre31 mm31 mm31 mm0
Mouth corner24 mm24 mm24 mm0
Jaw corner52.5 mm52.5 mm52.5 mm0
A reading of zero is the starting point, not a claim about your face. A face carries a left-right difference of its own, and this table says nothing about how large yours is. What it does is set a floor: whatever a test reports on a photo taken at an angle includes a camera term, and on a level face that term is the whole of the reading.

The camera term, written in millimetres

When a head turns, it turns about a vertical axis roughly through its middle. One side comes toward the lens and the other recedes, and the distance each side travels is the lever arm multiplied by the sine of the turn. At small angles the sine is close to the angle itself, which is what makes a two- or three-degree turn easy to dismiss and easy to photograph.

What a head turn does to the two sides
Head turnOne side closer, the other furtherAs a share of the 118 mm face width
0°0 mm0
2°3.3 mm0.028
5°8.3 mm0.070
10°16.5 mm0.140

The column that matters is the middle one, because depth is what a flat photo turns into apparent width. The nearer cheek is projected a little larger and the further one a little smaller, and the midline the scorer fits ends up sitting off centre for reasons that were never on the face. Two degrees is not a pose. It is the difference between facing the lens and almost facing it, and it is worth 3.3 mm before anything else is counted.

The same arithmetic runs the other way and is more useful in practice. Take the difference a test reports and ask how big a turn would have produced it with the face left alone.

The turn a reading implies
Reported left-right differenceTurn that produces it on its own
1 mm0.6°
2 mm1.2°
5 mm3.0°
8.3 mm5.0°
16.5 mm10.0°
Read the table backwards before you read your score. A 2 mm difference is inside a turn of about one degree — the amount you get from resting a phone on a shelf and leaning back. That does not make the difference unreal. It makes it unattributable until the photo is repeated at the same angle.

Do the test in three passes

1

Make a clean baseline

Use the rear camera when you can. Put the lens at eye level, step back, face it squarely, relax the expression, and pull hair off the outline. Use even front light, not a window on one side.

2

Compare only pairs

Look at paired features: eye level, brow height, nose position, mouth corners, and the broad outline at the jaw. Do not turn one feature into an overall verdict about attractiveness.

3

Retake before deciding

Take a second photo at the same distance, height and light. Ask one narrow question: does the same left-right difference appear again, at the same size?

The second photo is the test. A single image cannot separate the face from the pose, because both are in the same set of numbers. A repeat at a matched distance can, and it costs one more press of the shutter.

Where the reading stops

Four places to put the number down. Stop when the two photos were taken at different distances. The pose term scales with the lever arm and the lens distance, so two photos taken from different places are two different measurements, not two samples of one. Stop when the difference is smaller than the turn you allowed. A 2 mm reading sits inside a turn of about one degree; a 5 mm reading sits inside three. If the head was not held, the number is inside the pose. Stop when the tool does not show its midline. The midline is where the whole reading comes from, and a score with no line on it cannot be checked against the photo it came from. Stop and get the change looked at when it is new, one-sided, painful, or it affects speech, chewing or the eye. A landmark score is a reading of a photograph and is not the instrument for that.
Download the readings

The fixed head, the four level pairs, the camera term for turns up to ten degrees, the turn each difference implies, and the stop rules, as a CSV.

Download face-symmetry-test-chart.csv

The same face, turned

A level face with four matched left and right pairs sitting equal distances from the midline, beside the same face turned two degrees, where one side has come 3.3 mm closer to the lens and the other has receded by the same amount, so the fitted midline no longer sits on the middle of the face.
The same face and the same pairs, level and turned two degrees. Nothing on the face moved.

Watch the photo repeat check

Prefer YouTube? Open the video in YouTube.

Questions

How do I test my face symmetry at home?

Take two straight-on, neutral photos at eye level under even light, from the same spot. In each photo, compare matched features on either side of the midline. Treat a difference as a repeatable pattern only when it appears in both captures at about the same size.

What is a good face symmetry score?

There is no universal figure. A score reflects one tool's landmarks, midline and calculation in one photo, and the tools do not share a scale. It is more useful for comparing your own matched photos than for ranking yourself against anyone else.

Can a small head turn really change the result?

Yes, and the size of the change can be worked out. On a head about 190 mm front to back, a two-degree turn moves one side 3.3 mm closer to the lens and the other 3.3 mm further away, which the photo records as an apparent width difference. Five degrees is worth about 8.3 mm, and ten degrees about 16.5 mm.

Why do two apps give me different symmetry scores?

Because they compare different pairs. One may score the eyes, brows, nose, mouth and outline; another may weight them differently or fit its midline from a different set of anchors. The percentages are not interchangeable, and neither is a percentile.

Can a face symmetry test diagnose anything?

No. It compares visible positions in a photograph. A change that is new, one-sided, painful, or that affects speech, chewing or the eye needs a clinician, not a landmark score.

Related guides