How the Harmony Score works.

1. 01
   Detect. MediaPipe identifies a dense face mesh from a single photo on the device.
2. 02
   Measure. GLOWD selects anatomically useful points and calculates distances such as cheek width, facial height, and eye spacing.
3. 03
   Normalize. Distances become ratios, so the analysis is based on proportion rather than pixel size.
4. 04
   Score. Each ratio is compared with a configured reference center using a smooth peaked curve.
5. 05
   Explain. The result pairs the score with the measured value, reference context, percentile, and practical ideas.

A measurement receives its highest score at the configured center. The score decreases smoothly as distance from that center grows. There is no broad plateau that labels every value in a range as identical.

Scores are constrained to the app scale. Facial symmetry is treated differently: the measurement represents left-to-right deviation, so lower deviation scores better. Selected structural ratios carry modestly different weights in the combined Harmony Score.

GLOWD uses statistical reference distributions derived from a facial landmark dataset to provide percentile context. Reference centers and spread values are calibrated for the app's MediaPipe measurements. A percentile describes where a measurement sits in that reference distribution. It is not a ranking of a person's worth or attractiveness.

The top facial third uses an estimated hairline point because a standard face mesh does not directly observe the real hairline. Hair coverage can therefore influence that result more than ratios based on visible eye, nose, mouth, jaw, and cheek points.

Facial measurements from a photo are sensitive to pose, lens perspective, expression, light, shadows, and occlusion. For a more repeatable scan, face the camera directly, use a neutral expression, keep the full face visible, and choose even light.

GLOWD is an entertainment and educational tool. It does not diagnose health conditions, provide medical advice, or establish an objective standard of beauty.

Face photos and landmark analysis remain on the iPhone and are not uploaded for facial processing. Saved scan history is held in local app storage. GLOWD uses limited analytics, attribution, and subscription infrastructure as described in its privacy policy.