How to take a photo that gives an accurate facial analysis

Here is the uncomfortable truth about every facial analysis, ours included: it measures a 2D photo, and a 2D photo can lie. Before you read anything into a score, make sure the photo is not the thing being measured. Five habits do most of the work.

Square to the camera, head level

A small turn or tilt quietly changes width ratios, symmetry and every angle. Look straight down the lens with your head level. Our pipeline corrects for minor roll, but it cannot rebuild a face shot from three-quarters — start with a true front-on frame.

Soft, even, front light

Side light carves shadows that software can mistake for bone structure; overhead light hides the jaw and brow. The most reliable light is soft and pointed at you — a window on an overcast day is close to ideal. Avoid harsh single-source lighting that splits your face into bright and dark halves.

Distance is the silent distortion

This is the big one. A phone held close exaggerates whatever is nearest the lens — usually the nose — and shrinks the ears, which throws off width ratios and nose metrics in particular. Hold the camera at eye level and at least an arm's length away, or better, have someone shoot you from a step or two back. The same face can look like a different person at 20cm versus 1.5m.

Neutral face, clear hairline

Smiling moves the mouth, eyes and the entire lower third, so keep a relaxed, neutral expression. Pull hair off your forehead so the hairline and face outline are visible — and since the hairline is the one landmark no model can find, you will be asked to confirm it anyway. Lose the glasses if you can.

Do these and your numbers describe your face instead of your setup. A good test: take two photos in different conditions. If the scores swing a lot, the lighting and angle were the variable, not your face — re-shoot clean and trust that one.