Image-model photo score
A locally running image model evaluates five aligned views of the same face. Their mean is compared with 3,704 reference-photo values and mapped to a 60–99 score. The scale is specific to this model, not a percentage or global population rank.
Separate geometric descriptions
Face shape, eye angles and ratios use face landmarks. They do not explain or add up to the image-model score. Face shape matches are approximate templates, not calibrated probabilities.
Photo checks
We check face count, size, pose, sharpness, exposure and visible expression. Poor inputs require another photo. Automatic checks do not detect every filter or obstruction.
Haircut suggestions
A deterministic 50-style catalog ranks ideas using outline, proportions, selected texture and style preference. No hairstyle images are generated. Styling quality has not been established by user studies.
Comparisons and limits
Compare photos of the same consenting adult. Angle, light and expression affect the score. Public benchmark agreement and implementation consistency do not establish accuracy for every person or device.
Dedicated measurement tools
Golden ratio and thirds tools retain inspectable landmarks and optional point correction. Thirds require you to mark the visible hairline. These tools report measurements only, with no separate attractiveness rating.