What movement adds that framing cannot
A static frame gives you a composition. A moving frame gives you a sequence of compositions and a sense of the space that connects them. That is the real function of camera movements: they deliver information across time. A push in narrows attention. A pull out reveals context that was there all along. A track alongside someone tells you the walk matters more than the destination.
Movement also carries physical weight. A dolly on track feels intentional and expensive because a machine had to be built for it. Handheld feels contingent, present, human. A crane feels like an authorial hand lifting away from the story. Audiences do not name any of this, but they feel who is in control of the frame.
The standard moves
- Pan. Rotation around a vertical axis. The camera does not travel, so there is no parallax.
- Tilt. Rotation around a horizontal axis, up or down. Often paired with a pan on a fluid head.
- Dolly in and dolly out. The camera travels toward or away from the subject. Sometimes called a push in and a pull out.
- Truck or track. Lateral travel, usually parallel to a moving subject.
- Arc or orbit. Travel on a curve around the subject, which keeps the subject centred while the background rotates behind them.
- Crane or jib. Vertical travel, often combined with a tilt to hold the subject as the camera rises.
- Handheld. Operator-carried, with visible micro-instability that reads as immediacy.
- Steadicam or gimbal. Stabilised operator movement, so the camera can travel anywhere a person can walk without the shake.
- Zoom. Not a movement of the camera at all, but a change of focal length. Grouped here because it reads as one.
- Dolly zoom. A dolly and a zoom in opposite directions at once, holding subject size while the background warps.
- Whip pan. A pan fast enough to smear into motion blur, usually used as a transition.
- Roll. Rotation around the lens axis, which produces a dutch angle if you stop partway.
Real shots combine these freely. A single take might crane down into a track that ends on a slow push in, and shot lists write that as one line.
Motivating a move
The most useful question about any camera movement is what starts it and what stops it. A move that begins when a character stands and settles when they sit feels invisible. The same move over a static subject calls attention to the camera. Neither is wrong, but the second one is a statement and should be made on purpose.
Speed does most of the emotional work. A push so slow the audience cannot consciously detect it builds pressure across a whole scene. The identical move at four times the speed reads as aggression or comedy. When a move is not landing, the fix is more often the speed than the path.
Camera movements in AI video
This is where the vocabulary meets a real constraint. Models generate a few seconds at a time and have no persistent 3D model of the scene, so long or compound moves accumulate error:
- Reliable:
slow dolly in,slow push out,gentle orbit around the subject,static camera on a tripod - Risky: multi-part moves in one prompt, such as
crane down then track left then push in - Frequently ignored: precise speeds and distances, for example
dolly forward exactly two metres
Three practical habits. Name one move per clip and chain clips together for anything compound, using the last frame of one as the first frame of the next. State whether the subject moves as well, because "camera tracks left" with no subject cue often produces a subject that slides sideways too. And when a clip needs to hold absolutely still, say locked off and remove every verb of travel from the prompt, since drift is the default in most training footage.