10 AI Video Prompt Tips That Actually Work

Most AI video prompts are vague. "A cat" gets you a random cat. "A fluffy orange tabby cat stretching lazily in golden afternoon sunlight, slow motion" gets you the video you actually want.

Here are 10 techniques that consistently produce better results.

1. The SPEC Framework

Every great prompt has four elements:

• Subject: WHO/WHAT is in the video
• Place: WHERE (setting, background)
• Event: WHAT HAPPENS (action, motion)
• Cinematography: HOW it's filmed (camera, lighting, style)

Example:

> ❌ "A woman in a coffee shop"

>

> ✅ "A young woman in a vintage apron (S) behind a rustic wooden counter (P) carefully pouring steamed milk into a ceramic cup in a heart pattern (E), warm morning light from a window, shallow depth of field, 4K cinematic (C)"

2. Be Specific About Motion

Static descriptions produce static videos. Always describe what moves.

Examples:

• ❌ "A city street"
• ✅ "A bustling city street at rush hour, cars zooming past, people walking briskly, neon signs flickering"

3. Use Cinematography Terms

AI models trained on film understand camera language:

| Term | What it means |

|------|---------------|

| Wide shot | Full scene visible |

| Close-up | Face/object fills frame |

| Tracking shot | Camera follows subject |

| Dolly zoom | Background distorts |

| Shallow DOF | Background blurred |

| Low angle | Camera looking up |

| Bird's eye | Looking straight down |

| Handheld | Slight camera shake |

| Steadicam | Smooth tracking |

| Slow motion | Real-time stretched |

Example:

> "Slow motion close-up of water droplets falling onto a leaf, shallow depth of field, 4K cinematic"

4. Specify Lighting

Lighting changes everything. Always include it.

• Time of day: "Golden hour", "Blue hour", "Harsh midday sun", "Overcast"
• Source: "Window light from the left", "Backlit", "Neon glow", "Candlelight"
• Mood: "Warm and cozy", "Cold and sterile", "Dramatic chiaroscuro"

5. Reference Real Films/Photographers

Models know famous styles:

• "Wes Anderson symmetry"
• "Roger Deakins cinematography"
• "Blade Runner 2049 aesthetic"
• "Studio Ghibli style"
• "Vintage 70s film grain"

6. Avoid Negatives

Models struggle with "no" / "without" / "don't".

Examples:

• ❌ "A dog without a collar" → might still have a collar
• ✅ "A dog with no accessories" → better, but still imperfect
• ✅ "A clean-coated dog running on a beach" → describe what you want, not what you don't

7. Specify Aspect Ratio

Match the platform:

• 16:9: YouTube, website heroes
• 9:16: TikTok, Instagram Reels, Shorts
• 1:1: Instagram feed
• 4:5: Instagram feed (vertical)

8. Add Texture Details

Texture makes videos feel real:

• "Skin pores visible"
• "Fabric texture in detail"
• "Water droplets on glass"
• "Skin sheen from sweat"
• "Hair flyaways in wind"

9. Control Pacing

| Speed | Words to use |

|-------|--------------|

| Slow | "slow motion", "lingering", "suspended", "majestic" |

| Normal | (default) |

| Fast | "rapid", "energetic", "frenetic", "whip pan" |

10. Iterate with Specific Changes

First generation rarely perfects. Better to iterate:

• ❌ "Make it better"
• ✅ "The cat should be looking at the camera, not away"
• ✅ "Make the lighting warmer, more golden hour"
• ✅ "Add motion blur to the background"

Copy-Paste Templates

Product Demo

```

[Product name] on a clean white marble surface, soft studio lighting,

slow 360° rotation, shallow depth of field, 4K cinematic

```

Lifestyle

```

[Person description] enjoying [activity] in [location], natural lighting,

candid handheld camera, warm color grading, 4K

```

Social Media Reel (9:16)

```

Close-up of [subject], fast cuts every 0.5s, energetic music sync,

vibrant colors, vertical 9:16 format, TikTok-style

```

Food / Cooking

```

[Chef/dish description], steam rising, close-up macro shot, shallow

depth of field, warm kitchen lighting, slow zoom, 4K

```

Travel / Nature

```

[Location] at [time of day], sweeping drone shot, golden hour

lighting, gentle camera movement, 4K cinematic

```

Common Mistakes

1. Too vague: "Cool video" → useless
2. Too long: 200+ words → model loses focus
3. Contradictions: "Fast and slow motion" → incoherent
4. Generic subjects: "A person" → random person
5. Forgetting style: No aesthetic direction → default look

FAQ

Q: How long should prompts be?

A: 30-80 words is the sweet spot. Less than 20 = too vague. More than 100 = diluted.

Q: Do I need to mention the camera?

A: For best results, yes. Even simple "4K" or "cinematic" helps.

Q: Why does my video not match the prompt?

A: Models are probabilistic. Be specific. Iterate. Try rephrasing.

Q: Can I use negative prompts?

A: Limited support. Better to describe what you want.

Try these prompts on getaivideo →