Facial symmetry, measured rather than scored

What is actually being measured

Eleven pairs of landmarks — eye corners, eyelids, brows, nostrils, mouth corners, cheekbones, jaw angles and temples — each compared against its mirror. For every pair the tool reports two numbers: how much further one side sits from the midline than the other, and how much higher one sits than the other.

Those are the two ways a pair can be asymmetrical, and separating them is useful. A brow that is higher on one side is a different observation from a brow that is set further out.

Why the jaw and cheekbones are kept separate

Not every landmark is equally trustworthy in a photograph, and the difference is large enough that averaging them together produces nonsense.

Eye corners, eyelids, nostrils and mouth corners sit close to the flat front of the face. Cheekbones, jaw angles and temples sit four or five centimetres further forward of the axis your head turns about. When you rotate even slightly, a paired feature shifts by roughly twice its depth times the sine of the angle — so at eight degrees of turn the eye corners separate by about 3 mm while the cheekbones separate by about 14 mm.

That is not asymmetry. It is the camera. An earlier version of this tool averaged all eleven pairs into one figure, which meant a barely-noticeable head turn could manufacture two centimetres of apparent unevenness in a perfectly symmetrical face.

So the headline figure comes from the central features alone, and every pair — central or contour — now carries an explicit head-turn allowance. The tool measures your yaw, works out how much horizontal difference that angle alone would produce at that feature's depth, and subtracts it. Both the raw figure and the adjusted one are shown, so you can see how much of the result was the camera.

The allowance is deliberately generous, which means it can under-report genuine asymmetry. That is the safer direction of error for a tool nobody should be alarmed by. It is not a substitute for a square-on photograph: if the allowance column is doing most of the work, retake rather than trust the adjustment.

A smile is not asymmetry

Most people smile lopsidedly. One side of the mouth lifts further than the other, by a few millimetres, every single time — it is a habit of muscle recruitment rather than a property of the skull underneath.

That matters here because the mouth corners are one of the eight central features, and in a smiling photograph they are routinely the largest difference on the page. Counting them makes anyone who smiled for the camera look measurably less symmetrical than they are.

So the tool reads the expression directly from the face model's blendshape scores — and specifically reads how uneven the smile is, not whether one is present. A symmetric smile lifts both corners together and leaves the pair perfectly usable. Only when the two sides differ, or the grin is broad enough to distort the mouth regardless, are the mouth corners dropped from the headline figure and greyed out, with the measured smile asymmetry reported so you can see what was set aside.

For a structural reading, use a relaxed, neutral expression — lips together, no smile, jaw unclenched.

The midline is fitted, not assumed

The common failure in symmetry tools is measuring against the vertical edge of the photograph. Tilt your head five degrees and every paired feature on one side moves up while the other moves down, so the tool reports dramatic asymmetry that is entirely an artefact of how you held your head.

Here the midline is a least-squares line fitted through five landmarks that genuinely lie on the centre of the face: forehead, nose bridge, nose tip, lip centre and chin. Every measurement is then taken in that line's coordinate frame. A tilted head simply tilts the midline with it, and the numbers do not move.

What the tool cannot undo is perspective. A head turned toward or away from the camera genuinely foreshortens one side, and no midline fit recovers that. The tool checks head pose and warns you when the turn is large enough to matter.

Why there is no score

Almost every symmetry tool on the web returns a percentage. It is a strange thing to do, because the percentage carries no information the millimetres do not, while adding a strong implication that higher is better and that there is a number you ought to be reaching.

There is not. Measured asymmetry of a few millimetres is the ordinary condition of human faces, documented repeatedly in facial morphology research, and it is part of what makes a face recognisable rather than a flaw in it. Tools that mirror one half of your face to show a "perfect" version are generating something that has never existed, and most people find the result faintly unsettling for exactly that reason.

So this reports geometry and stops there. If your median deviation is 2.4 mm, that is the measurement. It is not a grade.

Reading your numbers sensibly

Take two or three photographs in different sittings before drawing any conclusion. If the median moves by more than a millimetre between them, the variation is coming from your camera and your posture rather than your face, and the honest answer is that a single photograph cannot measure this precisely.

Phone cameras are particularly unhelpful at close range. A wide-angle lens held at arm's length enlarges whatever is nearest it, so a slight turn is magnified into visible asymmetry. Standing further back and zooming in, or using a rear camera, gives a considerably more honest result.

When asymmetry is a medical question

Lifelong mild asymmetry is normal and is not what this section is about.

A sudden change in facial symmetry is a different matter. New drooping on one side, particularly with weakness, numbness, a change in speech, or an inability to close one eye, warrants urgent medical attention rather than a web tool. That combination can indicate a stroke or a facial nerve palsy, and both are time-critical. If you are reading this because something changed recently, stop and seek care now.

This tool measures geometry in a photograph on your device. It does not diagnose anything, expresses no view on appearance, and cannot substitute for medical assessment. Results depend heavily on camera position and head pose.

Questions people ask

Why is there no symmetry score out of 100?

Because a single score invites a verdict, and there is nothing to pass or fail. What is actually measurable is how far each paired feature sits from its mirror, in millimetres, so that is what this reports. A percentage would add no information and a great deal of implied judgement.

How symmetrical is a normal face?

Measurably asymmetrical, in essentially everyone. Studies of facial morphology consistently find differences of a few millimetres between the two sides in the general population. One to three millimetres of median deviation is the ordinary human condition, and it is invisible in conversation.

Why did my result change between two photos?

Almost certainly the photograph rather than your face. A head turned a few degrees, a camera above or below eye level, or a phone's wide-angle lens held close all create apparent asymmetry. The tool corrects for head tilt by fitting a midline, and it keeps the perspective-sensitive outer contour out of the headline figure, but it cannot undo perspective entirely. If two photos disagree, trust neither and take a third square-on.

Does the tool correct for a tilted head?

Yes. Everything is measured against a midline fitted through your forehead, nose bridge, nose tip, lip centre and chin, rather than against the vertical edge of the image. A tilted head therefore does not register as asymmetry, which is the most common way these tools produce nonsense.

Why were my mouth corners excluded?

Because you were smiling. Most people's smiles are lopsided by a few millimetres, which is a muscle habit rather than a feature of the face, and in a smiling photograph the mouth corners are usually the largest difference measured. The tool reads the expression from the face model and drops the mouth from the headline figure when a smile is detected. Retake with a neutral expression for a structural reading.

Should I be worried about asymmetry?

Not on the basis of a photograph. Mild asymmetry is normal and lifelong. A sudden change in facial symmetry, particularly if it comes with drooping, weakness, numbness or difficulty speaking, is a different matter entirely and needs urgent medical attention rather than a web tool.

Does this mean one side of my face is better?

No, and we do not offer that reading. Tools that mirror one half of a face to show you a symmetrical version are producing something that has never existed and generally looks wrong, because familiar faces are asymmetrical ones. The measurements here describe geometry, not quality.