AI Video Prompt Engineering Guide
AI Video Prompt Engineering Guide
TL;DR
- Every prompt has four parts: subject, action, environment, and style
- Be specific: describe details, not generic concepts
- Use 2-4 sentences for optimal results
- Think like a cinematographer: use camera and lighting terms
- Test and iterate to refine your prompts
The Four Building Blocks
1. Subject
What: The main focus of your video - person, object, or scene.
How: Describe appearance, clothing, position, and characteristics.
Example: "A woman in her 30s with long dark hair, wearing a flowing red dress"
2. Action
What: How things move in your scene.
How: Specify direction, speed, and motion style.
Example: "slowly spinning in a circle, her dress swirling around her"
3. Environment
What: The setting and atmosphere.
How: Include lighting, time of day, weather, and location.
Example: "in a sun-drenched meadow at golden hour, soft backlighting"
4. Style
What: The visual aesthetic and mood.
How: Use cinematography terms, film references, or artistic styles.
Example: "shot on 35mm film with shallow depth of field, cinematic"
Common Mistakes
| ❌ Bad | ✅ Good |
|---|---|
| "A cat playing" | "An orange tabby cat batting at a red yarn ball on a wooden floor, sunlight streaming through a window" |
| "A person walking" | "A hiker in outdoor gear walking along a rocky mountain trail at sunrise, mist rolling through the valley" |
| "Futuristic city" | "Slow aerial dolly over a neon-lit cyberpunk metropolis, flying cars between holographic billboards, rain-slicked streets" |
Why These Fail:
- Too vague: AI can't visualize "playing" or "walking" without context
- Missing details: No environment, no camera direction, no style
- No specificity: What kind of cat? What colors? What lighting?
Quick Tips
Length
Keep prompts to 2-4 sentences. Longer prompts dilute focus.
Camera Language
Use professional terms:
- "Slow dolly forward" instead of "zoom in"
- "Low angle looking up" instead of "shot from below"
- "Shallow depth of field" instead of "blurry background"
Test and Iterate
- Generate multiple variations
- Note what works and what doesn't
- Refine based on results
Example Prompts
Example 1: Wildlife Scene
Before: "A bird in a tree"
After: "A vibrant blue jay perched on a snow-covered pine branch, head tilting curiously. Shot in shallow focus with soft morning light filtering through frosted needles."
Example 2: Sci-Fi Environment
Before: "Futuristic city at night"
After: "Slow aerial dolly over a neon-lit cyberpunk metropolis at night, flying cars streaking between towering holographic billboards. Rain-slicked streets reflecting pink and blue neon glow."
Example 3: Nature Closeup
Before: "A flower blooming"
After: "Extreme closeup of a pink cherry blossom slowly opening at dawn, dew drops glistening on delicate petals. Macro lens with shallow depth of field, soft diffused backlight."
Conclusion
Great AI video prompts combine specificity and structure. Always think about:
- Subject: What's in the frame?
- Action: How does it move?
- Environment: Where and when?
- Style: What's the look and feel?
The difference between "a sunset" and "golden hour sunlight streaming through windswept grass, shot with shallow depth of field" is the difference between generic and stunning.
Ready to create? Try Seedance and put these techniques into practice.