Who do I look like?

An archetype is a stable pattern across the 17 structural metrics. Two people with very different individual features can still match the same archetype because the underlying ratios cluster. Strong cheekbone projection, a 5 degree positive canthal tilt, and a low FWHR around 1.7 land in one cluster; a deep-set brow, low gonial angle, and high FWHR land in another. The archetype is the cluster, not the face.

This matters because the most useful thing a look-alike tool can do is not guess a celebrity name. It is to tell you which structural traits are doing the heavy lifting in your face so you can stop trying to fix the ones that are already strong. Most users discover they have one elite metric, one weak metric, and fifteen unremarkable ones. The elite metric is the lever.

We do not surface celebrity photos because the legal ground around using a real person's likeness as a comparison anchor is genuinely contested in most jurisdictions. Right-of-publicity statutes vary by state and country. We label your archetype with descriptors (initials, descriptors, or score pattern only) and let you decide whether to search the label yourself. The score pattern is the real signal.

Step one: facial landmark detection. A TensorFlow.js model derived from the iBUG 300-W 68-point landmark spec locates 68 points on your face. The corners of your eyes, the tip of your nose, the philtrum borders, the nostril wings, the lip outline, the jawline curve from gonion to gonion, and the upper and lower forehead.

Step two: metric extraction. The 68 points feed 17 ratio calculations. Examples include canthal tilt (angle between the medial and lateral canthus), FWHR (cheekbone-to-cheekbone width divided by upper-lip-to-eyebrow height), facial thirds (forehead height divided by midface height divided by lower-face height), and gonial angle (angle of the mandible at the ramus-to-body intersection). Each ratio is normalized against published anthropometry norms.

Step three: percentile mapping. Each metric is mapped to a 0 to 100 percentile against the published norms. Step four: archetype classification. The 17 percentile values are clustered against archetype score patterns derived from published anthropometry research. The closest cluster is your primary archetype; the second-closest is your secondary archetype.

The output is descriptive. Your archetype is not a verdict. It is the cluster of metrics your face happens to fall into, and the report tells you which specific metrics are driving the classification so you can act on them.