Face symmetry test
See how your left and right sides line up, feature by feature, from one straight-on photo. Add a second photo to see how much of the difference is just the photo. Free, and your photo never leaves your browser.
Measured in your browser. Your photo is never uploaded.
How it's measured.
The face model places 24 pairs of matching points on the eyes, brows, nose, mouth, cheeks and jaw. Forma fits the centre line that best matches the pairs, mirrors each left-side point across it, and measures the gap to its partner. Gaps are a percentage of the distance between your pupils, so photo size doesn't matter. Each region's gaps are averaged, and the asymmetry index is the average of the five regions.
We tested the method's limits before publishing it:
- Head turn. We rotated the face model's own perfectly symmetric reference face. Each degree of turn added about 1 point to the index; nodding and tilting the head added nothing. So the test asks for a retake when your head is turned 5° or more, and warns from 2°.
- The model's own noise. We measured an AI-generated portrait and copies of it made perfectly symmetric by mirroring one half. The mirrored copies still scored about 0.8–0.9, and single regions up to about 1.6. That's why readings below 1.5 are reported as symmetric, and region values below 2 as noise. One test face isn't a validation; your own photo-to-photo spread is the better guide.
- Light and expression. Light from one side and a lopsided smile both shift points. The photo check flags them.
Cheek and jaw points sit on the edge of the face, where turn and shadow move them most. Read that region with the most caution.
What studies report.
Symmetry is one of the most studied facial traits, and the findings are more modest than the internet suggests.
- A small to moderate link, not a rule. A review across many face samples found natural symmetry and rated attractiveness were related at about r = 0.23, with wide variation between studies; averageness showed a stronger relationship (Rhodes 2006).
- Sometimes no effect at all. With naturalistic ratings, a 2019 study found no effect of symmetry while averageness still mattered (Jones & Jaeger 2019).
- Perfect isn't preferred. Faces made perfectly symmetric by mirroring one half were mostly not preferred to the originals (Kowner 1996).
- People only notice so much. In a study of 50 faces judged by 50 raters, perceived symmetry changed mainly over a narrow middle range of measured asymmetry (Eißing 2024).
That's why Forma reports where your sides differ and never converts it into a score of how you look.
Forma's asymmetry index is its own 2D definition and is not comparable with the indices in these studies. Sources checked 2 October 2026.
Questions.
How does this face symmetry test work?
Google's MediaPipe face model finds 24 matching left and right points (eyes, brows, nose, mouth, cheeks and jaw) in your browser. Forma fits a centre line to your face, mirrors each left-side point across it, and measures how far it lands from its right-side partner. The average gap, as a percentage of the distance between your pupils, is the asymmetry index.
What is a good face symmetry score?
There isn't one. The index describes one photo, and no face is a perfect mirror image. In our test, even a perfectly mirrored photo scored about 1 because of the face model's own placement error, so readings below 1.5 are reported as symmetric within what the method can detect.
Does facial symmetry make you more attractive?
Only a little, and not simply. A 2006 review found natural symmetry and attractiveness were weakly to moderately related, while averageness was a stronger factor. A 2019 study with naturalistic ratings found no symmetry effect. Faces made perfectly symmetric by mirroring one half were often rated less attractive. Forma does not turn symmetry into a beauty score.
Why does my score change between photos?
Head turn is the biggest cause: on our face-model test, each degree of turn added about 1 point of apparent asymmetry. Side lighting, expression and the model's own placement error add more. Add two or three photos and the tool shows your spread, so you can see which differences are noise.
Why do mirrored or flipped selfies look strange?
You usually see yourself in a mirror and others see the unflipped view. Small natural asymmetries switch sides when an image flips, so a familiar face can look slightly off. That is normal and is not something this test measures.
Is my photo uploaded?
No. The photo is decoded and measured in your browser tab and cleared when you leave or reload. Forma only counts that the tool was used, without any identifier.
Measure the rest the same way.
Width, eye spacing, lower face and side profile, with a photo check and one change to try.
Start a free report