What an attractiveness test can and can’t measure
An AI attractiveness test can compare visible facial measurements, but a harmony score does not capture how people respond to your appearance. Research on facial attractiveness has examined several measurable features:
- Averageness. Some studies find that faces blended from several photographs receive higher ratings than many of the individual faces.
- Symmetry. Studies have found links between symmetry and attractiveness ratings, although making a face artificially symmetrical does not consistently improve its ratings.
- Proportions. Experiments with facial spacing show that changing proportions can affect ratings. They do not establish a single formula for an attractive face.
This tool measures proportion and symmetry from one front-facing photo. It does not assess expression in motion, voice, personal style or the context in which someone sees you.
We call the result a facial harmony score. It compares measurements with the reference values described below. The research provides context for those measurements; it does not validate this tool as a predictor of attractiveness.
How the harmony score works
The tool finds 478 landmarks in your browser, accounts for sideways head tilt using your pupils, and estimates your facial midline. It scores five components from 0 to 100 and combines them with fixed weights:
| Component | Weight |
|---|---|
| Symmetry | 30% |
| Facial thirds | 25% |
| Facial fifths | 20% |
| Eye area | 15% |
| Golden ratio | 10% |
Higher scores mean a closer match to the tool’s reference values. These values and weights are settings of the tool, not population averages or a research-validated formula for beauty.
For ratio-based measurements such as the middle-to-lower third and lip proportions, the score peaks at the reference and decreases as the difference grows. A measurement one tolerance away scores about 82; two tolerances away scores about 46. Symmetry and fifths use a different curve that decreases as mismatch grows. The eye-area score combines tilt, spacing and symmetry.
Symmetry (30%)
We reflect matching landmarks on the left and right sides of the face across your midline and measure how far each pair is from lining up. This is done for four regions (eyes, brows, midface and the outline of the face) and averaged. The mismatch is expressed as a fraction of the distance between your pupils, so it doesn’t depend on photo size. A score of about 37 means the average mismatch is 15% of that distance.
- Landmark
- Mirrored from the other side
- Fitted midline
The score describes alignment in this photo. Head angle, expression and landmark estimates can all affect it. Research on symmetry and attractiveness does not make a symmetry score an overall rating of a face.
Facial thirds (25%)
The classical canon divides the face into three equal vertical sections: hairline to brow, brow to the base of the nose, and the base of the nose to the chin. Our landmarks can’t see the hairline, so we estimate the upper third and show it for reference only. The score compares the middle third with the lower third, with equal lengths (a ratio of 1.0) as the reference and a tolerance of 0.1.
Facial fifths (20%)
The fifths canon divides the width of the face at eye level into five equal parts, each one eye wide: from the side of the face to the outer corner of the eye, across the eye, the gap between the eyes, the other eye, and out to the other side. We measure all five widths, divide each by your average eye width and score how far, on average, they are from one eye width.
Eye area (15%)
Three measurements, averaged:
- Canthal tilt, the angle from the inner to the outer corner of each eye. This tool uses an upward tilt of 4° as its reference, with a tolerance of 4°. That reference is a scoring choice, not a requirement for an attractive face.
- Eye spacing, the gap between the inner corners compared with the width of one eye. The reference is equal (a ratio of 1.0).
- Eye symmetry, how closely the two eyes mirror each other.
Golden ratio (10%)
The golden ratio is φ ≈ 1.618. This component compares two measurements with it: mouth width to nose width, and lower lip height to upper lip height. A difference of 20% from φ gives a score of about 82 for that ratio; the two scores are then averaged.
The golden ratio is not an established universal rule for facial beauty. In one experiment on facial spacing, preferred eye-to-mouth distance was about 36% of face length, and preferred eye spacing about 46% of face width. Those findings concern specific images and measurements; they do not validate our mouth-to-nose and lip comparisons. The 10% weight used here is a product choice.
Reading your score
The overall score falls into one of five bands that describe the match to the reference values. These are labels used by this tool; they are not population percentiles or grades of attractiveness.
| Band | Score |
|---|---|
| Very close match | 85–100 |
| Close match | 78–84 |
| Moderate match | 68–77 |
| Limited match | 55–67 |
| Low match | 0–54 |
100 is the highest possible score. It means the measurements closely match the scoring references, not that the face is the most attractive. The relationship between proportions and preference is more complex: one study found that a blend of highly rated faces was preferred to a blend of all faces, even when its differences from the average were exaggerated.
We do not have a representative population average for this score. Free results are computed on your device and are not sent to us for statistics.
Under the result, highest-scoring area and lowest-scoring area identify the regions with the highest and lowest scores in that photo. They do not identify which feature attracts the most attention or which one needs changing.
Limitations
- One photo, front-facing. We measure a flat image. We can’t see your profile, depth, or how your face moves.
- Classical canons are idealizations. When thirds and fifths were measured on real young adults, most faces did not match the canons exactly. International studies also find typical proportions differ between populations.
- The references have limits. Some comparisons use classical facial canons and the golden ratio. The same reference values are applied regardless of sex or ancestry, and they cannot represent every population or preference.
- Landmarks are estimates. The model infers where each point sits, and soft or hidden features, such as a chin under a beard, are harder to place.
The component breakdown is more useful for understanding the result than treating the total as an overall judgment of your appearance.
How to get a fair measurement
To make results easier to compare, keep the photo conditions similar:
- Face the camera straight on, with your head level and your chin neither raised nor tucked.
- Keep a neutral expression. A smile changes the lower third, the lips and the eye area.
- Stand back. Close-up phone cameras exaggerate the center of the face. Hold the phone at arm’s length or have someone take the photo from a couple of meters away.
- Use even, front-facing light, such as a window in front of you, rather than a lamp from one side.
- Pull your hair off your face and remove glasses.
If two photos give different scores, compare head position, expression, camera distance and lighting before drawing a conclusion. Averaging the scores has not been validated as a more accurate result for this tool.
Sources
- Langlois, J. H., & Roggman, L. A. (1990). Attractive faces are only average. Psychological Science, 1(2), 115–121.
- Langlois, J. H., Kalakanis, L., Rubenstein, A. J., Larson, A., Hallam, M., & Smoot, M. (2000). Maxims or myths of beauty? A meta-analytic and theoretical review. Psychological Bulletin, 126(3), 390–423.
- Cunningham, M. R., Roberts, A. R., Barbee, A. P., Druen, P. B., & Wu, C.-H. (1995). “Their ideas of beauty are, on the whole, the same as ours”: Consistency and variability in the cross-cultural perception of female physical attractiveness. Journal of Personality and Social Psychology, 68(2), 261–279.
- Rhodes, G. (2006). The evolutionary psychology of facial beauty. Annual Review of Psychology, 57, 199–226.
- Grammer, K., & Thornhill, R. (1994). Human (Homo sapiens) facial attractiveness and sexual selection: The role of symmetry and averageness. Journal of Comparative Psychology, 108(3), 233–242.
- Rhodes, G., Proffitt, F., Grady, J. M., & Sumich, A. (1998). Facial symmetry and the perception of beauty. Psychonomic Bulletin & Review, 5(4), 659–669.
- Swaddle, J. P., & Cuthill, I. C. (1995). Asymmetry and human facial attractiveness: Symmetry may not always be beautiful. Proceedings of the Royal Society B, 261(1360), 111–116.
- Perrett, D. I., May, K. A., & Yoshikawa, S. (1994). Facial shape and judgements of female attractiveness. Nature, 368, 239–242.
- Farkas, L. G., Hreczko, T. A., Kolar, J. C., & Munro, I. R. (1985). Vertical and horizontal proportions of the face in young adult North American Caucasians: Revision of neoclassical canons. Plastic and Reconstructive Surgery, 75(3), 328–338.
- Farkas, L. G., Katic, M. J., Forrest, C. R., et al. (2005). International anthropometric study of facial morphology in various ethnic groups/races. Journal of Craniofacial Surgery, 16(4), 615–646.
- Holland, E. (2008). Marquardt’s Phi mask: Pitfalls of relying on fashion models and the golden ratio to describe a beautiful face. Aesthetic Plastic Surgery, 32(2), 200–208.
- Pallett, P. M., Link, S., & Lee, K. (2010). New “golden” ratios for facial beauty. Vision Research, 50(2), 149–154.
- Kartynnik, Y., Ablavatski, A., Grishchenko, I., & Grundmann, M. (2019). Real-time facial surface geometry from monocular video on mobile GPUs. arXiv:1907.06724.
