AI Attractiveness Test: Rate My Face 1–100

A geometry-based facial harmony score from 1 to 100, broken down by region, with the styling moves that flatter your strongest features.

Free no signup478 landmarksOn-device no upload~3 s to result

Live scan demo0 / 478 points

Front-facing · neutral expression · even light

Add a selfie to begin

Method

How it works and the exact formula

Most 'rate my face' apps return a single number without saying how it was produced. This attractiveness test does the opposite: it computes a 1–100 facial harmony score from 11 geometric measurements, shows each measurement and its weight, and breaks the result down into five regions (eyes, proportions, jaw and chin, nose and mouth, symmetry) so you can see where your score comes from.

Everything runs in your browser. MediaPipe Face Landmarker places 478 landmarks on your face; we convert them to pixel space and rotate the face so the line between your irises is level. We then measure bilateral symmetry, facial thirds balance, canthal tilt, length-to-width ratio, midface ratio, facial width-to-height ratio, eye spacing, jaw-to-cheek ratio, nose width, mouth-to-nose ratio and jaw angle.

Each measurement is compared with a reference target using a Gaussian closeness curve: a value exactly on target scores 100, one tolerance away scores about 61, two tolerances away about 14. Targets come from classical proportion canons and anthropometric literature, for example a mildly positive canthal tilt of about +5.5°, an fWHR around 1.9 and balanced thirds of 33% each. The eleven scores are combined with fixed weights, symmetry and thirds carrying the most, and mapped onto the 35–98 harmony scale.

It's important to be clear about what this measures. Attractiveness research finds that averageness, symmetry and sexual dimorphism all contribute to ratings, but so do skin, expression, grooming and familiarity, none of which a geometric scan captures. Treat the score as a description of proportions and a guide to styling, not a measure of worth.

To compare results fairly, keep conditions constant: same distance, same light, neutral expression, camera at eye level. A close selfie can lower your symmetry and proportion scores by several points because of perspective distortion, which has nothing to do with your face. If you scan several photos, the most frontal, evenly lit one is the most representative.

Formulas

Feature score: scoreᵢ = 100 · e^(−½ · ((valueᵢ − targetᵢ) ÷ toleranceᵢ)²)
Weighted sum: W = 0.20·symmetry + 0.14·thirds + 0.10·tilt + 0.10·LWR + 0.10·midface + 0.08·fWHR + 0.08·eye spacing + 0.07·jaw + 0.05·nose + 0.04·mouth + 0.04·jaw angle
Harmony: harmony = clamp(30 + 0.68 · W, 35, 98)

MediaPipe landmarks used

Anatomical point	Index	Role
Iris centers	468 / 473	Level the face (roll correction) before any ratio is computed.
Trichion (proxy) / glabella / subnasale / menton	10 / 168 / 2 / 152	Face length and vertical thirds.
Zygion	234 / 454	Cheekbone width: denominator for most ratios.
Gonion (proxy)	58 / 288	Jaw width and jaw angle.
Canthi	33 / 133 / 362 / 263	Canthal tilt and the gap between the eyes.
Alare / cheilion	102 / 331, 61 / 291	Nose and mouth widths.
12 bilateral pairs	e.g. 159 ↔ 386, 70 ↔ 300	Symmetry score.

Benchmarks

Facial harmony bands and how the scanner ranks them

Facial harmony bands and how the scanner ranks them
Metric	Band	Value range	Tool score	Modelled percentile
Harmony	Exceptional	88 – 98	88–98	94th–98th
Harmony	High	80 – 87	80–87	87th–93rd
Harmony	Above average	72 – 79	72–79	75th–85th
Harmony	Average	64 – 71	64–71	59th–73rd
Harmony	Below band	50 – 63	50–63	29th–57th

Harmony percentiles use the same modelled curve as individual features. Percentiles are modelled from the reference distribution our scorer uses (scores are mapped through a normal curve, mean 60, SD 18), not from a database of user faces. We don't build one: selfies are deleted within 2 hours and free scan results after 48 hours.

Styling

Styling and virtual try-on by result

Your regional breakdown tells you where styling has the most effect. Start with your lowest region.

Eyes strongest

Draw attention up: frames with a defined browline, groomed brows and face-framing layers that start at the cheekbone.

Find eye-framing glasses

Proportions lowest

Rebalance thirds with fringes (long upper third) or jaw-level volume (long lower third).

See your thirds and fifths

Jaw and chin lowest

Necklines and beards sit next to the jaw: V-necks and sharp collars add definition; tight fades sharpen the jaw corner.

Preview defining necklines

Symmetry lowest

Side parts and frames with a level top line make small asymmetries recede.

Run the symmetry test

Want the full picture

The full report renders 14 looks on your real photo and ranks them, with your color season and a styling plan.

See report plans

Attractiveness Test: frequently asked questions

Is a 75 face rating score considered attractive?

On our 35–98 harmony scale, 75 is above average: the scale is anchored so that about 70 is the midpoint of typical faces (PSL 4 on our PSL estimate), and 75 means most of your measured ratios sit inside their reference bands. A geometric score is not the same as how attractive people find you, which also depends on skin, expression and style.

How do I rate my face from 1 to 100?

Upload a front-facing, evenly lit selfie with a neutral expression. The test measures 11 ratios with 478 landmarks and combines them into a harmony score, with a breakdown by region.

Are AI attractiveness tests accurate?

They can be consistent, but most don't explain what they measure. This test is fully transparent: you can see every ratio, target and weight. What no AI test can do is measure attraction itself, which varies between people and cultures.

Why do different face rating apps give different scores?

They use different measurements, targets, weights and sometimes models trained on crowd ratings. Lighting, lens distance and head angle also change results. Compare scores from the same tool with similar photos.

Can I improve my attractiveness score?

Your bone structure is fixed, but a better photo (eye-level, arm's length, front light) often raises the score several points by removing distortion. Styling changes perceived harmony even when the measurement stays the same.

What makes a face attractive according to research?

Consistent findings include averageness (closeness to the population average), symmetry, sexual dimorphism, clear skin and positive expression. Rhodes (2006) and Little et al. (2011) review this research.

Is my photo uploaded?

No. The rating runs fully in your browser. Your photo is uploaded only if you opt in to the AI color analysis, and that copy is deleted within two hours.

Why is my attractiveness score lower than I expected?

Usually the photo: close range, a tilted head or uneven light can each cost several points. Retake it at arm's length at eye level. If one region still scores low, that's the region where styling helps most.

References

Rhodes G (2006). The evolutionary psychology of facial beauty. Annual Review of Psychology 57:199–226.
Little AC, Jones BC, DeBruine LM (2011). Facial attractiveness: evolutionary based research. Philosophical Transactions of the Royal Society B 366:1638–1659.
Thornhill R, Gangestad SW (1999). Facial attractiveness. Trends in Cognitive Sciences 3(12):452–460.
Farkas LG (1994). Anthropometry of the Head and Face (2nd ed.). Raven Press, New York.
Kartynnik Y, Ablavatski A, Grishchenko I, Grundmann M (2019). Real-time Facial Surface Geometry from Monocular Video on Mobile GPUs. CVPR Workshop on Computer Vision for AR/VR.

Keep exploring

Related free tools and guides

Free tool