Start here
How a scan works
A scan combines two independent reads of your face: geometry measured in your browser, and ratings from AI vision models. Here's each step.
The steps
- You upload two photos, front and side. They go straight to private storage through a one-time upload link.
- Your browser finds 478 landmarks on the front photo using Google's MediaPipe Face Landmarker. This runs on your device in about a tenth of a second.
- 25 front measurements are calculated from those landmarks: ratios like face width to height, eye spacing and facial thirds, angles like canthal tilt, and a symmetry measure.
- You place or confirm 14 points on your side profile (nose tip, chin, lips and so on). From them come 14 profile measurements with published clinical norms.
- Two AI vision models rate seven features from 1 to 10, independently and at the same time. Their ratings are averaged.
- Your overall score is calculated from those ratings, and the four pillar scores from ratings and measurements together.
Why two kinds of read?
Geometry is precise but blind to things like skin quality, and AI ratings capture overall impression but can drift. Using both, and showing you each one separately, means you can see whether a number comes from your proportions or from how a model perceived the photo.
| Measurements | AI ratings | |
|---|---|---|
| What it reads | Distances and angles between landmarks | Overall impression of each feature |
| Where it runs | In your browser | On our servers, through AI providers |
| Good at | Proportions, symmetry, profile angles | Skin, eyes, jawline definition, cheekbones |
| Sensitive to | Head angle, lens distance | Lighting, expression, image quality |
After the scan
Detailed AI assessments (56 items across three pillars), your improvement plan, simulations and Ask AI all build on the same scan. Nothing is re-measured unless you scan again, so results stay consistent while you explore.