How Overchat face analysis works

Two methods behind Overchat face tools: visual assessment by AI vision models and measurements from facial landmarks

Updated
On this page

In this article, you'll learn how Overchat AI analyzes faces, which facial measurements it uses, how attractiveness scores are calculated, and what scientific research supports these methods.

Two approaches to facial analysis

Overchat AI uses artificial intelligence and facial geometry to analyze facial features, proportions, symmetry, and attractiveness. Our facial analysis tools combine AI vision models with precise measurements derived from facial landmarks.

Overchat AI uses two complementary methods:

  • visual assessment
  • landmark-based facial geometry.

Visual assessment evaluates facial appearance and individual features, while geometric analysis measures facial proportions, symmetry, and feature positions.

Visual assessment

AI vision models examine facial features directly from photographs, evaluating their appearance, proportions, and relationships with other features.

This method is used for overall attractiveness ratings and individual feature evaluations. It also assesses characteristics such as facial contours and the overall balance of facial features.

Landmark-based facial geometry

Overchat AI uses a facial landmark detection model, a neural network that builds a 3D mesh of the face from a single photograph, to detect 478 facial landmarks. These landmarks represent defined positions around the eyes, nose, mouth, jaw, and facial outline.

Their coordinates are used to calculate distances, proportions, angles, and symmetry. The calculations follow fixed mathematical formulas, producing consistent measurements from identical landmark coordinates.

Geometric analysis adds quantitative information to visual assessments. For example, an AI vision model evaluates the appearance of the eyes, while landmark measurements determine their spacing, proportions, and orientation.

These measurements give the scoring process defined values to consider alongside the photograph. Their contribution to the overall score depends on the tool: Looksmaxing AI combines a geometry-based prediction with a visual score, while Rate My Face supplies available measurements to the scoring model.

Facial geometry measurements

Overchat AI measures facial proportions, symmetry, angles, and face shape using landmark coordinates.

Facial proportions

Facial proportions describe the relative dimensions and positions of facial features. They are expressed as ratios between distances measured using facial landmarks.

The analysis includes the following measurements:

MeasurementCalculation
Face length-to-width ratioFace height / cheekbone width
Jaw-to-cheekbone ratioJaw width / cheekbone width
Nose-to-face ratioNose width / cheekbone width
Eye spacing ratioDistance between inner eye corners / mean eye width
Eye aspect ratioEye opening height / eye width
Mouth-to-nose ratioMouth width / nose width

Using ratios allows facial proportions to be compared independently of image size.

The resulting values describe facial structure in a form the scoring model can use alongside its visual assessment.

Facial proportions play an important role in attractiveness perception. A study of facial feature spacing and attractiveness found that horizontal eye spacing and the vertical distance between the eyes and mouth influence attractiveness judgments.

Overchat AI uses these proportions to provide geometric information about facial structure and feature spacing.

Facial symmetry

Facial symmetry describes the correspondence between features on opposite sides of the face.

Overchat AI defines a facial midline using landmarks near the bridge of the nose and the chin. Landmarks on one side are reflected across this line and compared with corresponding points on the opposite side.

The calculation uses nine landmark pairs. The average distance between paired points is normalized by cheekbone width:

symmetry error
symmetry score
mean paired landmark distance
cheekbone width

A lower symmetry error indicates greater correspondence between paired landmarks. A score of 100 represents perfect correspondence according to the measured landmark pairs.

Facial symmetry is a well-studied characteristic of perceived attractiveness. Research reviews have documented associations between facial symmetry and attractiveness judgments.

Overchat AI uses symmetry measurements as one of the geometric characteristics considered in facial analysis.

The symmetry score contributes to the wider assessment alongside the other measurements.

Facial angles and vertical proportions

Facial angles describe the orientation of individual facial features.

One measurement is canthal tilt, the angle between the inner and outer corners of the eye. Overchat AI calculates this angle from landmark coordinates. A positive value indicates that the outer corner is higher than the inner corner in the photograph.

The analysis also measures three vertical facial regions: the upper, middle, and lower thirds.

Each region is expressed as a proportion of their combined height:

upper, middle and lower third
upper, middle and lower span
total facial span

These measurements describe the vertical proportions of the face and the distribution of facial features.

Face shape

Face shape analysis uses the relative widths of the forehead, cheekbones, jaw, and chin, together with an estimate of jaw angle.

Overchat AI compares these measurements with seven predefined face-shape templates. Each template contains reference values and tolerances for specific facial characteristics.

The geometric distance between the observed face and each template is calculated as:

distance between the observed face and a template
observed feature
template feature
feature tolerance

Smaller distances indicate a closer match between the measured proportions and a template.

The distances are converted into relative match weights, which are normalized across the seven templates. These weights indicate how closely the face matches each shape.

This comparison produces a face-shape classification. Facial attractiveness scores use a separate scoring process.

From facial analysis to scores

Overchat AI uses visual assessments and geometric measurements to generate facial attractiveness ratings. The scoring process varies depending on the tool.

Looksmaxing AI combines a visual model score with a geometry-based prediction when both are available:

overall score
visual model score
geometry model score

The result is rounded to one decimal place on a 10-point scale. When the geometry-based prediction is unavailable, the visual model score is used.

This approach combines the AI model's evaluation of facial appearance with numerical measurements of facial structure.

Rate My Face supplies the photographs and available geometric measurements to the scoring model. The model evaluates individual features and assesses the face as a whole to determine the overall rating.

Both tools evaluate facial characteristics through their respective scoring methods, with results expressed on a consistent numerical scale.

Scientific basis

Research on facial attractiveness has examined the effects of facial proportions, feature spacing, and symmetry on attractiveness judgments. These studies have identified measurable associations between facial characteristics and perceived attractiveness.

Research on shared and individual preferences in facial attractiveness has also identified common preferences for certain facial characteristics.

At Overchat AI, we use findings from these studies to inform the selection of facial characteristics used in our geometric analysis. This research provides a scientific foundation for examining facial proportions, symmetry, and feature positioning alongside AI-based visual assessments.

Applications in Overchat tools

Overchat AI applies visual assessment and facial geometry across four facial analysis tools.

ToolMethod
Rate My FaceEvaluates overall facial attractiveness and individual features using visual AI analysis, with geometric measurements providing additional information.
Looksmaxing AICombines visual assessment with geometric measurements to generate attractiveness ratings, feature evaluations, and personalized appearance suggestions.
AI Face Shape DetectorAnalyzes facial proportions and compares them with seven predefined face-shape templates.
Celebrity Look-AlikeUses AI vision models to identify facial resemblance to celebrities.

Each tool applies the relevant analysis methods to its specific task. Detailed scoring formulas, landmark definitions, and classification parameters are described in the methodology articles for individual tools.

Accuracy and limitations

The accuracy and consistency of facial analysis depend on photograph quality, camera perspective, head position, lighting, and facial expression.

For the most consistent geometric measurements, use a clear, frontal photograph with a neutral expression and unobstructed facial features.

Overchat AI analyzes facial landmarks detected in two-dimensional photographs. Landmark coordinates provide the basis for measuring facial proportions, angles, and symmetry, while AI vision models evaluate the visual appearance of facial features.

Geometric calculations use fixed mathematical formulas, ensuring reproducible results for identical landmark coordinates. Consistent photography conditions also help maintain comparable measurements across images.

All methodology topics