FREE · NO SIGN-UP · OFFLINE
Am I pretty? Eight proportion scores, honestly reported.
This test measures eight geometric ratios of your face (width to length, eye span, nose and mouth proportions, lip position), plus left-right symmetry, and compares each against reference values from classical proportion canons. It runs on your device: your photo is never uploaded, stored or shared.
Drop a front-facing photo here or tap to choose one.
Straight-on, neutral expression, hair off your forehead gives the most reliable result.
- Your photo never leaves your browser; the analysis runs on your device
- Three on-device samples averaged per photo to reduce single-shot noise
- Free, no sign-up, no watermark
What this score covers, and what it does not
This score covers facial proportions and left-right symmetry only. It does not cover changeable dimensions like skincare, hairstyle, teeth or makeup. No number here is a verdict on how you look or what you are worth. That boundary is not a disclaimer we bolted on afterwards; it is printed on every result because it is the honest scope of the method.
The eight ratios are compared against a synthetic reference face built from classical neoclassical canons (the same idea the early web face tests used), then each deviation is bucketed into a sub-score. Every card shows your ratio next to the reference value, so you can audit the method itself rather than take a number on faith.
How the scoring works
An on-device model (MediaPipe FaceLandmarker) maps 478 facial landmarks. From those points we compute eight dimensionless ratios (face width to length, lower to upper face, face width to eye span, face width to nose width, lip position within the lower face, mouth width to nose width, lip-to-chin balance, and face length to nose length), plus a four-pair symmetry deviation (outer eyes, inner eyes, mouth corners, nose wings). Each ratio's deviation from its reference value maps to a sub-score; the total is their mean with a small symmetry penalty, capped at 100. No image ever leaves this tab, and the reference values are constants, not a model trained on anyone's photos.
Attractiveness test FAQ
Is the attractiveness test accurate?
It is accurate at what it measures: eight geometric ratios and symmetry, compared against reference proportions. Research consistently finds that proportion measures capture only part of perceived attractiveness: skin, hair, expression, grooming and many changeable factors carry the rest. That is exactly why the result page prints the coverage boundary next to every score.
Is my photo uploaded to a server?
No. The detection model runs inside your browser on your own device. You can verify this yourself: open the Network tab of your browser's developer tools and watch. No upload request is ever made, and the page keeps working in airplane mode after the model loads.
Why do you average three samples?
Landmark detection on a single image can wobble by a pixel or two, which moves the ratios slightly. Running the measurement at three scales and averaging the sub-scores smooths out that noise. It is the same photo processed three times on your device: nothing extra is uploaded or stored.
What can actually change how my face looks?
A lot, and none of it is in this score: sun protection and skincare change skin texture; a haircut changes perceived face shape and proportions; teeth, brows, posture, sleep and expression all read into a face. The proportions this test measures are the stable part, which is why we report them as information, not as a verdict.
Is the score the same as other "am I pretty" tests?
The method family is the same (early sites like PrettyScale popularised ratio-based scoring), but we changed three things: automatic landmark detection instead of hand-placed dots, a visible reference value on every sub-score, and a hard-printed boundary statement about what proportions cannot tell you.