Face Centering AI

The overlay compares the detected face with the middle of the image. A small horizontal offset is useful as a preparation warning, but it is not an acceptance decision: authorities can assess pose, expression, lighting and identity details that a geometric check cannot judge.

Use the original camera file where possible. A screenshot, heavily compressed image or crop that cuts into the hair can make landmark detection less reliable. If the box is clearly misplaced, retake the photo against an uncluttered background before relying on the score.

Centring is only one of several geometric rules, and it is rarely the one that causes a rejection on its own. Most published specifications care more about head height — the distance from chin to crown as a proportion of the frame — and about the eye line sitting within a band measured from the bottom edge. A photo can be perfectly centred and still fail because the head fills too little of the frame, which usually happens when the original was taken from too far away and then cropped.

Treat the overlay as a preparation aid rather than a verdict. It reports what it can measure from landmarks in your image; it cannot see whether the lighting flattens your features, whether the expression is neutral enough, or whether the background shade matches the one your authority publishes. A photo that scores well here can still be returned, and a photo that scores slightly off-centre may still sit inside the tolerance the published rule allows.

One practical habit removes most centring problems before they start: stand further back than feels natural and crop afterwards. Phone cameras held at arm's length distort features and encourage a tilted, off-centre frame, while a photo taken from two metres away on a level camera leaves room to crop precisely to the head-height band your document asks for.