How do I control the camera in an AI video?
Updated October 27, 2026
Plan the move in Previs: pick it by name, such as push in, orbit, crane or dolly zoom, or describe it, and check it in 3D. Seedance 2.5 then makes the video with that exact move instead of guessing it from the prompt.
A prompt can name a move but not how far the camera travels, how fast, or what it keeps in frame, so each try is a new guess. Previs builds the move in 3D at real sizes first, and the video model takes it as its reference.
Moves you can name
- Basic: push in, pull out, truck, pan, tilt, pedestal, zoom, tracking, handheld and static. Up to three combine into one.
- Cinematic: circling (an orbit), a rising pull-back, an arc in for an entrance, a scenic slide, and the dolly zoom.
You can also say it in the chat: “slow push in on the dog” or “circle around them”.