1. Upload one clear photo
Front-facing, even lighting, no makeup, no filters, no glasses. The cleaner the photo, the more accurate your face rating.
How Attractive Am I? Free AI Attractiveness Test
Upload a photo and get an honest answer in about five seconds — your attractiveness score, face rating and face shape, calculated from 478 facial landmarks in your own browser. Free, no sign-up, no photo upload.
Formula shown → How it worksChoose or drop one photo here
JPG, PNG or WebP · Up to 10 MB / 20 megapixels
Free result in ~5 seconds · Photo stays on your device
Choose a photo to begin.
Read the formula
The Score in One Line
Four experimental geometry components. Each is clamped to 0–1; these references are product choices, not population percentiles.
score100 = round(1000 * (0.40*S + 0.20*T + 0.20*F + 0.20*G)) / 10; score10 = score100 / 10
The canonical score is 0–100, rounded once to one decimal. Divide that value by ten for the 0–10 display (two decimals). Component scores range from 0 to 1; bars are not population percentiles.
Before We Calculate
All x/y coordinates are converted to image pixels. The eye-center line is rotated to horizontal. The center of that line is the reference axis.
Coordinates are aligned to the eye axis in image pixels. Cheek width is d(234,454). Thirds use contour point 10, the midpoint of brow points 105/334, nose point 2 and chin point 152. Fifths use 234,33,133,362,263,454 along the aligned horizontal axis. Contour height is y152-y10. Point 10 is not the true hairline. FWHR uses cheek width divided by upper-lip-to-brow height; jaw angles are 2D contour angles.
1. Mirror similarity · 40%
Mirrored landmark pairs relative to face width.
S = clamp(1 - mirrored pair deviation / 0.12)
2. Facial thirds · 20%
Visible contour-to-brow, brow-to-nose and nose-to-chin proportions; not anatomical hairline thirds.
T = clamp(1 - sum(abs(visible third - 1/3)) / 0.60)
3. Facial fifths · 20%
Five horizontal intervals between cheek and eye landmarks.
F = clamp(1 - sum(abs(horizontal fifth - 0.20)) / 0.80)
4. Golden ratio · 20%
Contour height to cheek width compared with the model reference 1.618.
G = clamp(1 - abs(contour height / cheek width - 1.618) / 0.50)
Landmark Pairs We Use
33/263 · 133/362 · 159/386 · 145/374 · 105/334 · 98/327 · 61/291 · 234/454 · 172/397 · 103/332
Face Shape Reference Table
Nearest Euclidean reference using [contour height / cheek width, lower contour width / cheek width, upper contour width / cheek width], divided by tolerances [0.20, 0.15, 0.15]. These hand-set prototypes are exploratory labels, not validated classes.
| Shape | Contour / cheek width | Lower / cheek width | Upper / cheek width |
|---|---|---|---|
| Oval | 1.38 | 0.77 | 0.88 |
| Round | 1.12 | 0.85 | 0.91 |
| Square | 1.18 | 0.94 | 0.94 |
| Heart | 1.32 | 0.67 | 0.97 |
| Diamond | 1.40 | 0.72 | 0.77 |
| Oblong | 1.62 | 0.83 | 0.89 |
| Triangle | 1.30 | 0.97 | 0.80 |
When We Return No Score
No detected face, multiple faces, missing or non-finite required landmarks, out-of-frame required points and degenerate measurements produce an error. Small faces, near-edge framing, possible head pose changes and eye or mouth conditions may instead produce a result with a specific warning. The current pose check is a heuristic, not a measured 30-degree yaw threshold.
What We Do Not Measure
We do not measure objective beauty, skin quality, age, color season or anatomical hairline thirds. Contour angles are not bone angles. Photo perspective and model error affect measurements.
Algorithm Version & Reproducibility
harmony-0.3.0
face-landmarker-float16-v1 / tasks-vision-0.10.32
Compare the same photo with the same preprocessing, model and algorithm versions, browser, device and inference backend. GPU and CPU execution can produce small numerical differences. A version number alone does not guarantee identical scores across devices.
harmony-0.3.0 replaces eye width, centering and spacing components with visible thirds, horizontal fifths and contour height / cheek width deviation. Weights remain 40/20/20/20. Historical results retain their stored version.
How the Attractiveness Test Works
2. 478 landmarks become measurements
Your face is mapped into 478 points. From them we compute symmetry across both halves, visible facial thirds, horizontal fifths, canthal tilt, midface ratio, jaw angle and contour ratio deviation from the 1.618 reference.
3. You get a real score
Your attractiveness score (0–10) and face shape are free. The PDF report saves the measured geometry, scoring breakdown and the component with the largest weighted gap from its reference.
What Counts Toward Your Attractiveness Score
Symmetry (S, 40%)
Measures left-right differences across ten mirrored landmark pairs after eye-axis alignment. Small differences are normal.
Visible facial thirds (T, 20%)
Compares three vertical segments of the visible facial contour with equal thirds. The upper boundary is a mesh landmark, not the actual hairline.
Horizontal fifths (F, 20%)
Compares five horizontal segments, including eye widths and eye spacing, with equal fifths.
Contour ratio (G, 20%)
Compares facial contour height divided by cheekbone width with the product reference of 1.618. This reference is not a validated beauty standard.
How Attractive Am I? — Honest Answer
Attractiveness is not one number, and any site that gives you a single "out of 10" with no explanation is guessing. Your free score here is a weighted harmony index — it tells you how balanced your facial features are relative to each other. The PDF report explains each scoring component and its contribution. It does not predict changes to your appearance or provide treatment advice.
What this test is good for
Getting a baseline, tracking changes over months, and finding out which features are worth your attention.
What it is not
It is not a verdict on you as a person, and it is not a beauty pageant. Treat it as a fitness tracker for your face.
Frequently Asked Questions
Is this attractiveness test free?
Your free results include a harmony score, score band, highest-scoring component, component with the largest weighted gap from its reference, face-shape estimate and 22 measurements. No account is required and there is no usage limit. PDF reports are not yet available to purchase.
Is the score accurate?
The scoring formulas are deterministic for the same landmark coordinates and image dimensions. Landmark detection can vary with the model, preprocessing, device and browser. Photo quality, lighting and head position also affect measurements; identical photos are not guaranteed to produce identical scores across environments.
Do you upload or store my photo?
No photo is uploaded or stored. The model runs in your browser. Only an explicitly shared summary or the measurements you agree to submit for a PDF report are stored.
How is this different from a face rating app?
Everything is calculated locally from real measurements, every formula is shown, and you are not asked to make an account to see a real result.
Additional Descriptive Measurements
Eye area
Canthal tilt, eye spacing and eye width describe the eye area. Eye widths and spacing also feed the horizontal fifths component; there is no separate eye-area score.
Jaw and lower face
Jaw angle and jaw-to-cheekbone width ratio describe the lower face. Lower-third height also feeds the thirds component. Chin projection is not measured.
PDF measurement report
PDF reports are not yet available to purchase. The planned report saves the same 22 measurements, four scoring components, formulas and photo-quality notes. It does not add a scoring component or measure skin clarity.