This is the most controllable way to direct a generated shot. You are no longer describing motion in words; you are specifying where the shot begins and where it ends, and letting the model solve the middle. Composition, framing and continuity all become decisions you make in stills.
It is also how episodic work holds together. The last frame of one shot becomes the first frame of the next, so a sequence can run for minutes without the world quietly reorganising itself between cuts.
QUESTIONS PEOPLE ASK
How do I control motion in AI video?
Supply the first and last frame rather than describing the move. If the model only accepts one image, supply the first frame and keep the requested move to a single named camera action.