How FaceScan Calculates Your Beauty Score (and What It Actually Measures)

If you've used FaceScan's Beauty Score feature, you've seen a number out of 100 — plus breakdowns for eyes, nose, lips, jawline, forehead, and golden ratio proportions. But where does that number actually come from?

This post walks through the full methodology: the landmarks we measure, the formulas behind each sub-score, how they combine into an overall rating, and — importantly — what the score does not capture.

Everything runs on-device. No photo leaves your phone.

The Short Version

FaceScan uses ML Kit Face Mesh to detect 468 3-D facial landmarks from a single photo. From those points we calculate geometric ratios — eye width relative to face width, nose centering, jaw taper, facial thirds balance, and more.

Each ratio is compared to research-backed ideal values (with separate ideals for male and female faces). Six feature categories are scored 0–100, averaged, then passed through a power curve so the final score spreads out meaningfully across the population.

The result is a proportion score, not a judgment of personal attractiveness. Lighting, skin quality, hairstyle, expression, and taste are not part of the calculation.

What the Score Scale Means

These bands describe how closely your facial geometry aligns with established anthropometric ideals — not how "beautiful" you are in any absolute sense.

Step 1: Map Your Face with 468 Landmarks

Before any scoring happens, the app needs precise coordinates for key points on your face.

MediaPipe Face Mesh (via ML Kit) places 468 landmarks across your forehead, eyes, brows, nose, mouth, cheeks, and jaw. We project those 3-D points onto the 2-D image and measure distances between specific pairs.

Key landmarks include:

• Forehead and chin — top and bottom of the face for height measurements
• Cheek width points — widest part of the face
• Eye corners — inner and outer edges of each eye
• Nose tip and nostrils — width and horizontal position
• Mouth corners — lip width
• Jaw points — lower face taper

All of this happens locally on your Android device in under a second.

Step 2: Score Six Feature Categories

Each category produces its own 0–100 score. The overall Beauty Score is derived from all six.

1. Eyes (symmetry, size, spacing)

Three sub-scores, weighted and combined:

• Symmetry — how closely left and right eye widths match (ideal: perfect 1:1 ratio)
• Width ratio — average eye width as a fraction of total face width
• Gap ratio — distance between inner eye corners relative to eye width (the Rule of Fifths: the gap between your eyes should equal roughly one eye width)

Eye symmetry carries the highest weight (40%), reflecting research linking facial symmetry to perceived attractiveness.

2. Nose (width, proportion, centering)

• Width ratio — nose width relative to face width
• Nose-to-mouth ratio — how the nose width compares to mouth width
• Centering — how far the nose tip deviates from the vertical centre of the face (even small lateral shifts are penalised heavily)

3. Lips (fullness and proportion)

• Mouth-to-nose ratio — lip width relative to nose width
• Mouth-to-face ratio — lip width relative to total face width

Ideals differ by gender: female faces typically score against fuller lip proportions relative to the nose and face.

4. Jawline (face shape and taper)

• Face ratio (height ÷ width) — overall face shape (oval vs compact/square)
• Jaw taper — how much the jaw narrows below the cheekbones

Feminine ideals lean toward a slightly longer oval (h/w ≈ 1.45); masculine ideals toward a more compact proportion (≈ 1.30).

5. Forehead (facial thirds and brows)

The face is divided into three vertical zones:

Third 1 — hairline to brow midpoint
Third 2 — brow midpoint to nose tip
Third 3 — nose tip to chin

Each third is compared to an equal one-third of total face height. Deviation from balance lowers the score. Brow width relative to face width is also measured.

6. Golden Ratio (distinct proportions)

This category uses measurements not already covered by the other five features, to avoid double-penalising the same property:

• Upper/lower face ratio — distance from forehead to nose tip, divided by nose tip to chin (ideal: φ ≈ 1.618)
• Interpupillary ratio — distance between eye centres as a fraction of face width (ideal: 46%, from Pallett et al. 2010)
• Eye-to-mouth ratio — distance from eye centre to mouth centre as a fraction of face height (ideal: 36%, also from Pallett et al. 2010)

Step 3: Compare Each Measurement to an Ideal

Every sub-score uses the same penalty pattern:

subScore = clamp(100 − |actual − ideal| / ideal × penalty)

The further a measurement drifts from its research-backed ideal, the more points are deducted. Different properties use different penalty multipliers — nose centering, for example, is penalised more aggressively than width ratios, because even small lateral shifts are visually noticeable.

Each sub-score within a category is then combined using fixed weights (e.g. eye symmetry at 40%, width at 35%, gap at 25%).

Step 4: Combine Into the Overall Score

The six feature scores are averaged, then transformed with a power curve:

rawOverall = (eyes + nose + lips + jawline + forehead + goldenRatio) / 6
overall    = clamp( rawOverall^1.5 / 10 )

Why the power curve? Without it, most faces would cluster between 75 and 80 on individual features — because no single measurement is usually far from ideal. The curve spreads the distribution so the bands actually mean something:

Gender-Aware Ideals

All ideal ratios are gender-specific. FaceScan accepts "female", "male", or "neutral" (an average of both).

Gender affects ideals for:

• Eye width relative to face
• Nose width
• Lip fullness proportions
• Face height-to-width ratio
• Jaw taper
• Brow width

These norms come from anthropometric research — primarily Farkas (1994) and related orthodontic/aesthetic literature — not from subjective opinion.

What the Beauty Score Does NOT Measure

This is worth stating clearly:

• Skin texture, tone, or clarity — not measured
• Facial expression or mood — a smile changes landmark positions
• Hairstyle, makeup, or grooming — outside the geometry
• Perceived age — a separate feature in FaceScan
• Cultural beauty standards — ideals are derived from Western anthropometric research
• Subjective attractiveness — personal taste accounts for significant variance in real-world ratings

The score reflects geometric proportion relative to published ideals. Two people with identical scores can look completely different — and both can be genuinely attractive to different observers.

Tips for Accurate Results

1. Face the camera directly — head tilt and rotation shift landmark positions
2. Use even, frontal lighting — harsh shadows distort perceived proportions
3. Keep a neutral expression — smiling widens the mouth and raises the cheeks
4. Fill the frame — more pixels mean more precise landmark placement
5. Select the correct gender setting — ideals differ meaningfully between male and female norms

The Research Behind It

FaceScan's methodology draws on established literature in facial aesthetics and anthropometry:

• Pallett, Link & Lee (2010) — interpupillary and eye-to-mouth optima
• Farkas (1994) — canonical gender-specific facial proportion norms
• Rhodes (2006) — why symmetry correlates with attractiveness ratings
• Powell & Humphreys (1984) — the rule of facial thirds
• Ricketts (1982) — golden ratio applications in facial analysis

The approach is geometric and transparent — not a black-box neural network trained on crowd-sourced "hot or not" ratings.

Try It: FaceScan

FaceScan calculates your Beauty Score entirely on your Android device. Point your camera or pick a photo from your gallery — no uploads, no account, no cloud processing.

What you'll see:

• Overall Beauty Score out of 100 with interpretation bands
• Six feature breakdowns — eyes, nose, lips, jawline, forehead, golden ratio
• Individual sub-scores within each category

Related FaceScan features:

• Facial Symmetry — left/right balance with visual overlays
• Golden Ratio — dedicated φ analysis with grid overlay
• Face Shape — oval, round, square, heart, diamond, or oblong
• Face Mesh — 468-landmark visualisation
• Age Estimator — AI-based age range prediction
• Face Comparison — similarity scoring for two faces

The Takeaway

FaceScan's Beauty Score is a transparent, geometry-based proportion analysis — not a verdict on how you look. It maps 468 facial landmarks, compares six categories of ratios against research-backed ideals, and compresses the result into a 0–100 scale designed to spread meaningfully across real faces.

Use it to understand your facial proportions, track how lighting and pose affect measurements, or explore how geometric ideals relate to your features. Just remember: proportion is one lens. Character, expression, style, and personal taste matter far more in how a face is actually perceived.

All analysis in FaceScan runs on-device. Results depend on lighting, pose, and image quality, and are intended for informational and entertainment purposes.