Skip to main content

How do I write a good AI video prompt?

Use this guide to write video prompts that produce dynamic, accurate videos instead of flat or random results.

Write your video prompt like a short shot brief, in this order:

Subject + Motion + Environment + Camera + Style/Lighting + Audio (if supported)

The six numbered blocks below explain what to include in each part of your video prompt:

  1. Subject — who or what appears in the video, with specifics: age, look, clothing, how many. Avoid vague terms like "a person."

  1. Motion — clear action verbs, in the order they happen.

  1. Environment — place, time of day, weather. Never leave the setting unspecified.

  1. Camera — how the camera moves: dolly (the camera physically moves toward, away from, or alongside the subject), pan, orbit, handheld, tracking, close-up, wide. Camera direction is the most commonly skipped element and the number one cause of flat, static videos.

  1. Style/Lighting — for example: cinematic, natural light, film grain, anime, documentary. Pick one or two; don't stack many conflicting styles.

  1. Audio — ambient sounds, dialogue, or deliberate silence (only on video models that generate audio).

Example video prompt:

"A young woman in a red coat walks through a rainy Tokyo street at night, neon reflections on wet asphalt, slow dolly-in, shallow depth of field, soft rain and distant traffic."

Did this answer your question?