How to write AI video prompts that actually work
A simple five-part formula for prompts that give you the shot you imagined, with examples you can copy.
Great AI video starts with a clear prompt. The models are powerful, but they cannot read your mind. Here is the formula our team uses every day.
The five-part formula
- Subject. Who or what is on screen. Be specific about age, clothing and expression.
- Action. What happens during the clip. One main action reads better than three.
- Setting. Where and when. Time of day changes everything about the light.
- Camera. Describe the move: slow push-in, orbit, handheld follow, drone pull-back.
- Mood and style. Film stock, color palette, references like "warm golden hour" or "neon noir".
A before and after
Weak: a woman in a city.
Strong: a young woman in a yellow raincoat walks through a rain-soaked Tokyo alley at night, neon signs reflecting in puddles, slow tracking shot from behind, cinematic anamorphic look, moody teal and magenta palette.
Use the prompt enhancer
In the studio, switch on Prompt enhancement and the AI rewrites your idea with camera and lighting details. You can always edit the result before you generate.
Keep shots short and focused
Most models shine at 5 to 8 seconds. For longer stories, finish a clip and use Continue from last frame to chain the next shot, keeping the same character and setting.
Try it in the studio
Every technique in this article works in NovaMotions AI. See the credit price before you generate.
Keep reading
All articlesImage to video: turn any photo into a moving shot
Pick the right source image, describe the motion and use first and last frames for perfect transitions.
Credits explained: what a generation really costs
How credits work, why longer and higher-resolution clips cost more, and how to stretch your plan.