How to Structure an AI Video Prompt

Updated

An AI video prompt describes what happens over time: the subject, the action, the camera movement and the pacing, one shot per sentence. Unlike an image prompt, it must say how things move and in what order.

The four parts

  1. Subject and setting: who or what, and where (same as an image prompt).
  2. Action: what the subject does, with a clear start and end.
  3. Camera: the move (push-in, pan, orbit, handheld follow) and its speed.
  4. Pacing and style: duration, rhythm, look (commercial, documentary, anime).

One shot per sentence

Shot 1 (3s): close-up of hands kneading dough on a floured wooden table, slow push-in, warm morning light. Shot 2 (3s): the baker slides the loaf into a stone oven, camera at eye level, flames glowing. Shot 3 (4s): a golden baguette is torn open, steam rising, macro lens, shallow depth of field. Cinematic bakery commercial, natural sound.

Camera vocabulary

MoveWhat it does
Push-in / pull-outCamera moves toward or away from the subject
Pan / tiltCamera rotates left-right or up-down from a fixed point
Tracking / followCamera moves with the subject
OrbitCamera circles the subject
StaticCamera does not move; the action carries the shot

Starting from an image

Most video models can animate a still image. Generate or pick the first frame, then describe only the motion: what moves, how the camera moves, and how the shot ends. This gives far more control over the look than text alone.

Common mistakes

  • Too much action for the duration: a few seconds fit one clear movement.
  • No camera instruction: the model picks a slow drift.
  • Describing a state instead of a change: "a woman smiling" is an image; "a woman turns toward the camera and smiles" is a video.

Prompts to study

Real prompts from the network, each published with the result it produced:

More guides