If you've tried AI video generation, you've probably noticed that the same scene can feel amateur or cinematic depending on how the "camera" moves. Filmmakers have a shared vocabulary for this: dolly, pan, tilt, crane, orbit. AI video models have learned that vocabulary too.
Learning a dozen camera terms is one of the fastest ways to improve your AI videos. This cheat sheet explains each move, when to use it, and how to prompt for it. You can try the examples on Pixwit, which lets you pick a model and set camera direction alongside your prompt.
Why camera movement matters
Camera movement tells viewers where to look and how to feel. A slow push-in builds intimacy or tension. A pull-back reveals scale. A static shot lets the action speak. Without direction, AI models often pick a generic gentle drift, which looks fine but forgettable.
The essential moves
1. Static (locked-off)
What it is: The camera doesn't move at all. Use it for: Dialogue, product details, letting a strong action play out, loops. Prompt: "static camera, locked-off shot"
Static shot of a candle flame flickering in a dark room, smoke curling upward.
2. Push-in (dolly-in)
What it is: The camera moves physically toward the subject. Use it for: Building emotion, emphasizing a reveal, product hero shots. Prompt: "slow push-in," "dolly-in toward her face"
Slow push-in on an old man's face as he looks out a rain-streaked window.
3. Pull-back (dolly-out)
What it is: The camera moves away, revealing more of the scene. Use it for: Reveals, endings, showing scale or isolation. Prompt: "slow pull-back revealing…"
Slow pull-back from a lone hiker to reveal a vast snowy mountain range.
4. Pan
What it is: The camera rotates left or right from a fixed position. Use it for: Scanning a landscape, following a horizontal action. Prompt: "slow pan left across…"
Slow pan right across a busy night market, lanterns glowing.
5. Tilt
What it is: The camera rotates up or down from a fixed position. Use it for: Revealing height, such as buildings, trees, or a character from feet to face. Prompt: "tilt up from … to …"
Tilt up from the base of a redwood tree to its canopy, sunlight streaming through.
6. Tracking shot (follow)
What it is: The camera moves alongside or behind a moving subject. Use it for: Action, walking scenes, sports, vehicles. Prompt: "tracking shot following…," "camera follows from behind"
Tracking shot following a cyclist racing down a coastal road at sunset.
7. Orbit (arc shot)
What it is: The camera circles around the subject. Use it for: Products, heroic character moments, showing a subject in 3D. Prompt: "slow 360-degree orbit around…"
Slow orbit around a vintage motorcycle in a dim garage, light glinting off chrome.
Pixwit's Bullet Time effect is a packaged take on this idea: according to its page, it produces a 12-second fashion-film-style clip with a freeze-frame and a 360° orbital camera move, built from your photo plus optional reference images for the location and props.
8. Crane / jib up
What it is: The camera rises vertically, often while looking down. Use it for: Establishing shots, grand finales. Prompt: "crane up revealing…"
Crane up from a street café to reveal the rooftops of Paris at dusk.
9. Drone / aerial
What it is: A high, sweeping shot from above. Use it for: Landscapes, travel, real estate, openers. Prompt: "aerial drone shot gliding over…"
Aerial drone shot gliding over turquoise lagoons and white sand islands.
10. Handheld
What it is: Slightly shaky, organic movement. Use it for: Realism, documentary feel, social-native content. Prompt: "handheld camera, slight natural shake"
Handheld shot of friends laughing around a bonfire on the beach at night.
11. Zoom (vs. dolly)
What it is: The lens changes focal length; the camera doesn't move. Perspective stays flat. Use it for: Retro looks, dramatic snap zooms, comedy. Prompt: "fast snap zoom onto…"
Fast snap zoom onto a surprised cat's face, comedic.
12. Dolly zoom (vertigo effect)
What it is: The camera moves toward the subject while zooming out, or the reverse, warping the background. Use it for: Shock, realization, suspense. Prompt: "dolly zoom, vertigo effect"
Dolly zoom on a woman in a hallway as she realizes something, background stretching, suspenseful.
Models don't always nail this one. Expect a few attempts.
Combining moves with pace and lens
Camera terms get stronger with modifiers:
Speed: "slow," "smooth," "fast," "whip"
Lens: "wide-angle," "telephoto," "macro"
Angle: "low angle," "high angle," "eye level," "overhead"
Low-angle, slow tracking shot of a knight walking through fog, wide-angle lens, epic.
Tips for reliable results
One camera move per clip. "Orbit while pushing in while tilting" usually fails.
Use built-in camera controls when the tool offers them, and keep the prompt consistent with them.
Match the move to the duration. A full 360° orbit in a few seconds can look rushed; try a 90–180° arc.
For image-to-video, leave space around the subject so pull-backs and orbits have room.
Compare models. Some follow camera instructions more precisely than others.
Quick reference table
| Goal | Move |
|---|---|
| Build emotion | Push-in |
| Reveal scale | Pull-back, crane up |
| Show a product | Orbit, slow push-in |
| Follow action | Tracking |
| Feel real | Handheld |
| Establish place | Drone, crane |
| Create shock | Dolly zoom, snap zoom |
| Let action speak | Static |
Conclusion
You don't need film school to direct AI video. You need a dozen camera words and a sense of when to use them. Save this cheat sheet, try one move per clip, and watch your results improve. To practice with several video models, visit https://pixwit.ai.
Comments
Join the discussionNo comments yet. Be the first to comment.