Motion tracking using two reference points — sufficient for flat surfaces and basic 2D movement. Faster than four-point, but vulnerable to rotation.
With two-point tracking, you work with the minimum amount of information that still functions reliably for flat surfaces. You mark two distinct, high-contrast points in the image—ideally with maximum distance between them—and the tracker calculates position and scaling from these. This is sufficient for straight wall inserts, monitors, posters, or windows, as long as the surface remains parallel to the camera and no rotation comes into play.
The practical advantage lies in speed: two points track faster than four, the computer has less to calculate, and with stable shots—such as locked-off or minimally moving camera—the result is achieved on the first try. You save time in batch processing when you have to feed hundreds of shots with simple inserts. But here lies the trap: as soon as the camera rotates or the plane tilts, the two-point tracker loses orientation. A slight camera rotation leads to visible sliding or distortion of the insert—this can be corrected afterward, but then you should have opted for four points from the start.
In practice, I mainly use two points for quick proof-of-concept tests or when the scene's geometry is absolutely guaranteed to remain static. A typical scenario: a letter lying on a table, and you want to composite an image onto it—the camera doesn't move around the table, it stays frontal. Then it works elegantly. But as soon as I have a feeling that even minimal perspective shift is coming, I switch to four points. This is the professional habit—better to track for a minute longer and have the problem solved than to experience surprises later in conform. The two-point track is therefore not a cost-saving tool, but a tool for very clearly defined, flat scenes. Use it consciously, not out of convenience.