Motion tracking
Letting software follow a point or surface in your footage so graphics stick to it as the camera and subject move.
See it
What it is
Motion tracking analyzes footage, finds a feature it can recognize from frame to frame, and hands you back that feature's position over time. Parent a graphic to those numbers and it rides along with the shot. Three flavors matter. Point tracking follows features: one point gives you position only, and it takes two points before the solve can also work out scale and rotation from the distance and angle between them. Planar tracking (Mocha style) locks a shape to a flat surface like a wall, phone screen, or license plate, and can additionally solve the surface's perspective and deformation as it turns. 3D camera tracking, also called matchmoving, reconstructs where the camera actually moved so you can drop 3D objects into the scene.
Reach for it whenever something in frame moves and something you added has to move with it: floating labels over products, a censor blur that follows a face, screen replacement on a laptop, sparks stuck to a hand, or stabilization (same math, applied inverted).
Gotcha: the tracker needs a high-contrast feature that stays in frame, stays in focus, and never gets occluded. Heavy motion blur, rolling shutter wobble, low light, and someone walking in front of your track point all make the solve slide off. Pick a corner or a screw head, not a soft gradient, track a longer stretch than you need, and expect to hand-keyframe the frames the software gives up on.
Ask AI for it
Motion track this footage and attach a text label to the moving object. Use a two-point track on high-contrast features for position, scale, and rotation, or planar-track the visible surface if the object is a flat plane. Attach the label to the solved transform, and manually correct any occluded or drifting frames rather than letting the solve slide. Add motion blur to the graphic matching the footage, and hold the label 30 px above the tracked point with a thin leader line.