Start with the whole face
Measure from the central hairline to the bottom of the chin. Measure width between the widest cheekbone points. A hidden hairline makes the length uncertain; do not substitute the eyebrows.
Your selfie, seen through geometry.
Processed on this device.
Automatic landmarks run locally. Drop one portrait here or choose a photo.
JPG, PNG, WebP; HEIC if your browser can decode it. No file size cap; device memory limits apply.
Place the first endpoint, then the second.
Tap two endpoints. Keyboard: arrow keys move the crosshair; Shift moves faster; Enter places a point. All photo distances use pixels.
Enter direct measurements, or straight-line pixel distances from one front-facing photo. Keep your face level and expression relaxed. Do not mix photo pixels with physical units.
Measure from the central hairline to the bottom of the chin. Measure width between the widest cheekbone points. A hidden hairline makes the length uncertain; do not substitute the eyebrows.
Mouth width runs corner to corner; nose width runs between the outer nostril edges. Lower-face distances use the center line where the lips meet.
Use the center where the lips meet. Compare its straight-line distance to the chin with its straight-line distance to the base of the nose.
R1: hairline center H → chin 152, divided by cheekbone landmarks 234 → 454. The model estimates the visible cheekbone positions; correct them if needed.
R2: mouth corners 61 → 291, divided by outer nose wings 98 → 327.
R3: lip center L (midpoint of 13 and 14) → chin 152, divided by L → nose base 2.
All distances are straight lines in 2D pixels. Hair segmentation estimates H from a boundary above the forehead. If no boundary is found, H = chin + 1.15 × (mesh-top point 10 − chin). This fallback is a specified heuristic, not a measured hairline or scientific calibration. Drag H or another point, then press Update calculation.
distance = |ratio − 1.618| ÷ 1.618
component score = round(max(0, 100 − 250 × distance))
complete score = (score 1 + score 2 + score 3) ÷ 3
The 250 multiplier is a design choice, not a scientific constant or official standard. Distance is displayed as a percentage. The complete score is available only when all three components exist; missing components are named, never filled in.
This is not a beauty rating. This is not a population percentile. Photo measurements are estimates. Models and WASM are served from this local folder; no remote inference API is used. One image at a time; no login or result history.
This golden ratio face test compares your inputs with geometry. It does not measure attractiveness, health or measurement accuracy. Camera perspective and point placement can change the result; a calibrated error margin is unknown.
Reference definitions are informed by The Face Report and Golden Ratio Face. Different tools choose different landmarks, so their scores are not interchangeable. A 2024 research review by F. B. Naini found no convincing evidence linking phi to facial beauty. These references do not validate a beauty score.
No. It calculates ratios from selected distances. Closeness to phi does not establish how attractive a person is.
Yes. Fill in face length and width. Add optional measurement groups only when you have the required values. Use one unit throughout.
Landmarks, photo perspective, rounding and scoring formulas can differ. This page uses your stated measurements and the fixed comparison value 1.618; its combined score is a custom mathematical comparison, not a beauty rating.
No. They are explicitly synthetic arithmetic examples: face length 20 and width 13 in any consistent unit. They are not a person, population average or accuracy claim.