Design Openbook

An 8-Field Template for Video Prompts

How should you structure a prompt for AI video generation? Fill eight fields: subject, action, intent, place and time, camera, lighting and style, length, and what to avoid, and change only one field at a time when revising.

AI video looks off mostly because what, why and how far are missing. Set the fields in advance and fill them in for stable results.

How to use it

  • Split into fields: subject, action, intent, place and time, camera, lighting and style, length, and prohibitions.
  • Write the action as one specific verb. Not "plays happily" but "dives to catch the ball".
  • Set the range: how many seconds, how many shots, where it starts and where it ends.
  • Also write what must not happen. For example: no text on screen, no changing the face.
  • When fixing a result, change one field at a time. If you change several at once, you cannot tell what had the effect.

Prompts to copy

Blank template

[Subject]
[Action] (one verb, be specific)
[Intent] (the emotion this scene should convey)
[Place and time]
[Camera] (one movement name + height)
[Lighting and style]
[Length] 5 seconds, 1 shot
[Avoid] No text on screen, do not change the person's face

Filled example

[Subject] A seven-year-old child in a yellow raincoat
[Action] Jumps into a puddle with both feet together
[Intent] The playfulness of the moment after hesitating for fear of being scolded
[Place and time] An apartment playground on an afternoon after the rain
[Camera] Low angle, static. Slow motion only at the moment of the jump
[Lighting and style] Soft sunlight breaking through the clouds, photorealistic
[Length] 5 seconds, 1 shot
[Avoid] No text on screen

Source posts: @lowanglecuts

All cards (27)