Vizard Agent

How to cover a number plate with your own text that tracks the car

Last updated 2026-10-10 · 2 min read

Tell Vizard Agent what the plate should read instead and that nothing else in the frame gets touched. It measures where the plate sits in every frame as the car moves away, then composites the word at exactly that size and position, so it reads as part of the car rather than a patch over it.

What do you type into Vizard Agent?

Say "nothing else" and mean it literally. Requests to cover a plate are routinely answered with a general privacy pass that softens faces, other cars and half the street, and the clip comes back looking censored rather than clean. Then say what the plate should read and that it moves with the car.

Prompt

Cover the number plate on my car in this clip and make it read Koda instead. It should fit exactly where the plate is and move with the car as it drives away. Do not blur anything else in the shot: no faces, no other cars, nothing.

What does Vizard Agent actually do?

Almost all of the work is measurement rather than rendering, because the plate gets smaller and fainter in each frame as the car drives away into the haze. Vizard Agent checks every automatic box against the picture and projects the far positions it can't measure. In the session behind this article it:

When does this not work well?

Some shots will not take a fitted replacement. A plate at a sharp angle, one mostly hidden behind a tow bar or a bike rack, or footage so compressed that the plate is a grey blur gives nothing to measure, and a small blur is the honest answer there. Fast cuts and several cars each need their own tracking.

Common questions

Can it just blur the plate instead? Yes, if you prefer. Vizard Agent will, though a fitted word looks less hidden.

Can it do several cars? Yes. Vizard Agent treats each plate as its own tracked object.

How small can the plate get? Down to a few pixels. Beyond that, Vizard Agent projects the position.