Understand facial symmetry

Facial Symmetry

Facial symmetry is a relationship between corresponding features, not a single universal number. Start here to understand what a photo test measures and choose a guide about scores, camera effects, proportions or mirrored images.

In this guide
  1. What is facial symmetry?
  2. How does this facial symmetry test work?
  3. Choose the tool that fits your question
  4. Why can the same face look more symmetrical in another photo?
  5. Does facial symmetry determine attractiveness?
  6. Check your understanding
  7. Facial symmetry questions
  8. Start with the question you have
  9. Sources and further reading
CorrespondenceCompare the two sides

Ask where paired features fall relative to a defined reference line.

ProportionsCompare two lengths

Ask how a chosen height, width or distance relates to another.

Image effectsChange the view

Flip a picture or repeat one half to explore an artificial composite.

Check a photo’s symmetry Find your guide

What is facial symmetry?

For bilateral symmetry, imagine reflecting one side across a line and comparing it with the other. Matching points would occupy the same positions after reflection. In a photo, eye corners, brow points, nose edges, mouth corners and outline points provide specific places to compare.

The reference is part of the definition. A line fitted through several facial points can produce a different result from one defined by the eyes. Likewise, a few paired points describe less of a face than a full surface measurement. “Symmetrical” is incomplete unless you know which features, which reference and which image were assessed.

Facial asymmetry describes departures from that correspondence. A single photo score cannot identify their cause or separate anatomy from expression, camera projection and landmark estimation. For questions about a physical difference rather than a photo measurement, start with our facial asymmetry guide.

How does this facial symmetry test work?

The MediaPipe Face Landmarker estimates feature points in the photo. This site’s separate measurement code then:

  1. Defines each eye center as the midpoint of its inner and outer corners.
  2. Levels the line joining those centers and puts a vertical comparison line through their midpoint.
  3. Reflects one point of each selected pair across that line.
  4. Divides the straight-line gap to its counterpart by eye spacing and averages all 16 pairs.

The six groups are eyes, eyebrows, nose, mouth, cheeks and jaw. Each pair has equal weight; groups with more pairs contribute more to the average. These are estimated points in two dimensions, not measurements of bone shape, skin texture or facial depth.

Bilateral correspondence across an eye-defined reference line The eye centers define a horizontal line and its midpoint defines the vertical comparison line. A filled blue point on the left is reflected to a hollow blue point on the right. Its corresponding brown right point sits lower and farther outward. The connecting segment is the reflected mismatch, which contains both horizontal and vertical differences. Eye-defined reference Eye spacing Left point Reflected Right point Gap ÷ eye spacing = pair mismatch
Original schematic of one pair after eye-line leveling. The blue point is reflected across the dashed line; the short brown segment measures the gap to its partner. It combines sideways and vertical displacement. The checker averages 16 such pair measurements.

Measurement: mean reflected gap ÷ eye spacing.
Displayed mismatch: measurement × 100.
Unrounded index: 100 − mismatch percentage, with a floor of 0.

If the average gap is 3 px and eye spacing is 100 px, mismatch is 3% of eye spacing and the index is 97. That does not mean 97% of the face is symmetrical. Read the score guide for worked examples, rounding and regional weights, or inspect the full method and selected pairs.

Why the eye reference matters

Leveling the eye line absorbs the height difference between the two eye centers into the alignment. A small eye-region residual therefore cannot establish that the eye centers were level in the original photo. If you want to mark corner positions relative to that original image, use the manual eye symmetry test; it asks a different measurement question.

Dividing by eye spacing removes uniform image scale from the equation: doubling every coordinate doubles both the gap and the denominator. It does not correct a turned head, perspective or a misplaced landmark. Adding more mathematical factors would need validation for a stated purpose; complexity alone would not establish accuracy.

Choose the tool that fits your question

Related measurements and visual experiments answer different questions.
Your questionUseful starting pointWhat you get
Where do selected left and right features differ?Face symmetry checkerReflected pair mismatch and region overlays.
How do two photos compare?Two-photo comparisonTwo results with capture conditions to consider.
What does a particular feature show?Smile, nose or jaw photo testDefined manual points for a focused question.
How long or wide is one feature relative to another?Face proportion calculatorA ratio from two selected lengths.
What changes when I repeat one half?Mirror face analysisLeft-half and right-half composites.
Why does the reversed view look different?Inverted filterThe supplied image beside a horizontal flip.

A mirrored composite duplicates the same pixels, including lighting and expression, from one half. It is an invented image. A proportion such as face height ÷ width describes two lengths and can change independently of bilateral correspondence. The golden ratio guide explains why a mathematical ratio does not establish a facial ideal.

Why can the same face look more symmetrical in another photo?

A photograph projects a three-dimensional subject into two dimensions. Turning toward the camera changes where features project. Expression changes visible feature positions. Lighting and obstruction can affect which points the detector estimates.

Ward and colleagues’ photographic model shows how shorter camera distance changes apparent nasal width relative to facial width. It illustrates perspective, without providing a correction for this checker’s symmetry score. Read our selfie perspective guide for the distance demonstration and why scores change for capture and detector effects.

A useful photo experiment

Compare one deliberate change

  1. Choose a baseline. Use an eye-level frontal photo, a relaxed expression, even light and visible features.
  2. Write down the setup. Record the camera, distance, expression and lighting direction.
  3. Choose a single question. For example, compare a neutral expression with a smile while keeping the camera position and light the same.
  4. Inspect the relevant points. Note the region and raw mismatch in the comparison tool, alongside what you changed.

The conclusion concerns these photographs. A numerical difference alone does not demonstrate a lasting physical change. Keeping other conditions similar helps you interpret the experiment; it cannot eliminate all capture and detector variation.

Does facial symmetry determine attractiveness?

This checker has no validated attractiveness prediction. Research on perception uses particular images, raters and measurement methods. In Lee and colleagues’ 2025 study of 100 white adult face images, symmetry did not independently predict attractiveness ratings in the main analysis. That finding concerns the study’s sample and design, rather than every person or viewing situation.

The symmetry and attractiveness evidence guide covers the research question in more detail. To explore how duplicating one half changes an image, read perfectly symmetrical faces and mirror composites.

Check your understanding

Open each example to see why the measurement’s definition matters.

The same landmark coordinates are all doubled. Does the mismatch ratio double?

No. The reflected gaps and eye spacing both double, so their ratio stays the same. This is a property of the arithmetic for fixed points. Running a detector again on a changed image introduces a separate question about its estimates.

A repeated-half portrait looks balanced. Does it show the original face’s symmetry?

It shows pixels copied from one side. That construction cannot establish how the original two sides correspond. Compare the original landmarks if your question is bilateral geometry.

Two websites return different scores. Which is correct?

Their scores may define different quantities. Compare selected points, reference line, denominator, weights and score transform first. Determining accuracy requires an appropriate reference and validation of the intended measurement, rather than choosing the larger score.

Facial symmetry questions

How can I check my facial symmetry?

For a photo-based estimate, use a clear frontal image and inspect paired landmarks, the comparison line and regional residuals. This site publishes its 16-pair method. A photograph provides projected geometry; it cannot establish full three-dimensional anatomical symmetry.

Is there a normal facial symmetry percentage?

This checker has no representative population reference or validated normal range. Its mismatch percentage is relative to eye spacing; the score is a display index. A percentage from another app may use a different definition.

Does 100 mean a perfectly symmetrical face?

The headline rounds the index, so 100 can coexist with a small nonzero mismatch. Even an unrounded 100 would describe only the selected estimated pairs under this alignment. It would not establish that every surface, depth or movement matches.

Does flipping a photo change its symmetry?

An exact horizontal reflection preserves the reflected distances and eye spacing for corresponding transformed landmarks. The arithmetic is unchanged. A detector run again on the flipped pixels can estimate slightly different points, so independently generated scores need not be identical.

How accurate is this online test?

We have verified implementation behavior, but have not conducted a human accuracy study, clinical validation or representative population study for this score. Check whether the estimated points fit the image and read the methodology. Similar results across apps or repeated photos alone do not establish anatomical accuracy.

Explore the idea

Start with the question you have

Choose a path. Each guide answers a different question and connects to the relevant photo task.

Sources and further reading

The point selection, alignment and equation describe this site’s published method. The following sources provide detector, photography and research context; none validates this product’s score.

Continue with the question that brought you here:

Privacy settings

Photo analysis works without analytics. You can optionally allow Google Analytics to measure basic page visits. We do not send photos, filenames, face measurements or scores to analytics.

Analytics is off unless you allow it.

Your choice is stored in this browser. Turning analytics off after allowing it reloads this page; save any wanted result first. Read the privacy policy.

Check my face