How to write a video prompt that gives the scene direction
A practical guide to describing subjects, actions, camera movement, and lighting for your next AI video.
A good video prompt describes something the camera can see. Begin with one clear moment, then add enough detail to explain how that moment should unfold.
Build the scene around one action
Name the subject, the setting, and the action. “A chef tosses vegetables over a hot pan in a restaurant kitchen” gives the model a more concrete scene than “an amazing food video.” Keep the action achievable within the clip length you select.
Tell the camera where to go
Choose one camera direction: a slow push-in, a steady side track, or a locked-off close-up. Several conflicting camera instructions can make a short shot harder to control.
Describe light and atmosphere
Add a small number of visual details that support the scene. Warm window light, soft shadows, drifting steam, or wet pavement can establish a mood. Avoid adding details that compete with the main action.
Try a complete prompt
A ceramic cup on a wooden table beside a window. Steam rises slowly from the cup. The camera gently tracks sideways, with warm morning light catching the glaze. Keep the background quiet and softly out of focus.
Choose the settings before generating
In the homepage generator, choose Text to video and enter your prompt. Select a Veo model, then choose its available duration, resolution, and aspect ratio. Controls depend on the model, and the displayed credit cost changes with your choices. Sound is enabled for models that support it.
Review one thing at a time
Check whether the subject, movement, and composition match your intent. For another attempt, change one instruction so you can understand its effect. Each new generation uses the credits shown before submission. Download the result you want to keep; provider delivery links are not a permanent archive.