Zero-Upload Facial Symmetry, Golden Ratio (1.618) & Face-Shape Lab (2026)

Modern facial recognition pipelines detect 68 to 468 2D/3D nodal landmarks (such as interpupillary distance, nasal bridge width, and zygomatic arch curvature) to align and normalize a face before projecting it into a 128- or 512-dimensional vector embedding (like ArcFace or FaceNet) for cosine-similarity search.

Most smartphone front cameras use a wide-angle 23mm–26mm equivalent lens held only 12–15 inches from the face. Because the nose is ~1.5 inches closer to the lens than the cheekbones and ears, perspective foreshortening makes the nose appear up to 30% wider and narrows the bizygomatic width compared to an 85mm portrait taken from 5 feet away.

Key classical Phi (1.61803) proportions include: Total Face Length (Trichion to Menton) divided by Bizygomatic Cheekbone Width; Mouth Width (Cheilion to Cheilion) divided by Nose Width (Alare to Alare); and Eye-to-Mouth vertical distance divided by Eye-to-Nose-Tip distance.

First codified by Renaissance anatomists and plastic surgeons, the vertical face is divided into three equal segments (33.3% each): Upper Third (Trichion hairline to Glabella brow), Middle Third (Glabella brow to Subnasale base of nose), and Lower Third (Subnasale to Menton bottom of chin).

Never. The image is rendered directly inside a local HTML5 <canvas> element in your browser's RAM—zero bytes leave your device.