Camera moves for AI video: dolly, pan and more

A good AI video is not only about what is in the frame, but also about how the camera moves. A slow dolly toward your subject feels completely different from a static shot or a quick pan. Most video models understand camera language, as long as you actually use that language. Below you will learn which camera moves exist and how to write them into your prompt so the model performs them cleanly.

The main camera moves

Every move has a fixed name from the film world. Use that name literally in your prompt and the model knows what you mean.

• Dolly-in (push-in): the whole camera slides toward the subject. It builds tension or intimacy.
• Dolly-out (pull-out): the camera slides away instead, revealing the surroundings or creating a sense of distance.
• Pan: the camera pivots horizontally, left or right, while staying in place. Handy for showing an environment or following a moving subject.
• Tilt: the same pivot, but vertical, up or down. Great for conveying height or scale.
• Tracking (truck): the camera moves sideways along with the action, as if walking beside your subject.
• Crane or pedestal: the entire camera rises or lowers. Different from a tilt, where only the lens pivots.
• Orbit: the camera circles around the subject.
• Static (locked-off): the camera stays still and only the subject moves. Calm and clear, ideal for a conversation or a product shot.

A dolly is not a zoom

These two get mixed up a lot. A zoom changes the focal length of the lens, while a dolly physically moves the whole camera. They look similar, but a dolly gives real depth: the foreground and background shift relative to each other.

If you do not want your dolly to read as a flat zoom, name the depth in your prompt. Mention something in the foreground, something in the middle and something in the background, so the model can build the space.

Match the move to the feeling

A camera move tells part of the story, so do not pick one at random.

• Push-in for tension or an emotional moment.
• Pull-out to reveal context or make someone small in their surroundings.
• Tracking to add pace and energy to a walking or action scene.
• Static when all the attention should stay on the subject.

How to write it into your prompt

1. Name one move explicitly with the right term, for example "slow dolly-in toward the subject".
2. Keep the description short. A few words about the camera is enough; the rest of your prompt covers the scene.
3. Choose slow over fast. Calm moves come out much cleaner; fast moves quickly cause warping and blurry trailing artefacts.
4. Want several moves? Generate them as separate shots and place them one after another, instead of cramming everything into one prompt.

Common mistakes

• Two moves in one prompt. "Zoom in first and then pan right" rarely works. The model tries both and lands neither cleanly. Split it into two shots.
• Going too fast. A quick orbit or whip-pan sounds exciting but often produces artefacts. Keep moves calm, especially within a short clip.
• Vague camera language. "Nice camera movement" tells the model nothing. A concrete term like "tilt up" or "tracking left" does.

Not sure how to phrase a camera prompt? Let the prompt generator turn your idea into a full description, camera move included. Ready to try it? Create an account and bring your first shot to life with the video generator.