Prompting an AI video model is fundamentally different from prompting an image model. You're not describing what something looks like — you're describing what happens. Movement, time, and physics all matter.
The 4-Part Video Prompt Structure
Every strong AI video prompt answers four questions: What is the subject doing? What is the camera doing? What is the environment? What is the mood/aesthetic?
[Subject action] + [Camera movement] + [Environment/setting] + [Lighting/mood] + [Film style]Camera Language Is Everything
Most people describe what they want to see. Professionals describe how the camera sees it. These are different things entirely.
- Dolly in — camera physically moves toward subject (creates intimacy)
- Tracking shot — camera follows subject from the side
- Crane shot — camera rises or descends dramatically
- Steadicam — smooth handheld movement
- Rack focus — focus shifts from foreground to background
Add "slow motion, 240fps" to any action scene. AI video models handle high-speed subjects much better when you specify this.
Platform Differences Matter
Runway Gen-3, Kling, and Veo all have different strengths. Runway excels at cinematic quality and camera control. Kling handles complex character motion better. Veo produces the most natural physics.
Runway — Best for:
- Cinematic scenes with defined camera movements
- Slow-motion effects
- Dramatic lighting transitions
Kling — Best for:
- Character-driven scenes
- Complex hand and face movements
- Sports and action sequences
A Real Example Breakdown
An astronaut in a white spacesuit walks slowly across the rust-red surface of Mars. Cinematic wide shot, low camera angle, dust particles floating in the air, dramatic sunset on the horizon, God rays through the atmosphere, Steadicam movement, IMAX qualityThis prompt works because every element serves the camera: "low angle" creates scale, "Steadicam" defines movement quality, "God rays" adds atmospheric depth. Nothing is wasted.
