Writing great AI video prompts is part science, part art. After analyzing thousands of successful prompts and testing hundreds of our own, we've distilled the process into a repeatable formula that works across all major platforms.

🎯 The Core Rule: Be specific about WHAT moves, HOW it moves, and what the CAMERA is doing. These three elements determine 80% of your output quality.

The Perfect Prompt Formula

Here's the structure we recommend for every prompt:

[SUBJECT + ENVIRONMENT] + [MOTION DESCRIPTION] + [CAMERA MOVEMENT] + [LIGHTING/ATMOSPHERE] + [STYLE KEYWORDS]

Let's break down each component:

Step 1: Subject & Environment

Start by clearly describing what's in your scene. Even though the AI sees your image, reinforcing the subject description helps it understand what to animate.

  • ✅ "A woman with long brown hair standing in a sunflower field"
  • ✅ "Rocky ocean cliffs at sunset with crashing waves below"
  • ❌ "A person outside" (too vague)

Key principle: Your subject description should be 1-2 sentences maximum. Don't over-describe — the AI can see your image.

Step 2: Motion Keywords

This is the most critical part of your prompt. You must tell the AI exactly how objects in your scene should move.

Motion Words by Category

Gentle/Subtle: swaying, rippling, fluttering, drifting, floating, pulsing, breathing, shimmering

Flowing: flowing, cascading, pouring, streaming, rushing, rolling

Dynamic: crashing, exploding, bursting, spinning, rotating, swirling

Human/Animal: walking, running, turning, blinking, smiling, laughing, stretching

Speed modifiers: slowly, gently, rapidly, suddenly, gradually, smoothly

✅ Pro Tip: Stack motion words with speed modifiers for precision: "the flag gently fluttering and slowly rippling" gives the AI much more to work with than just "the flag moving."

Step 3: Camera Movement

Camera movement transforms a static animation into a cinematic experience. Here are the key camera keywords:

Camera MovePrompt KeywordEffect
Zoom In"slow zoom in", "push in"Creates intimacy and focus
Zoom Out"slow zoom out", "pull back"Reveals scope and context
Pan"pan left to right", "horizontal pan"Reveals wider scene
Tilt"tilt up", "tilt down"Creates reveal or scale
Dolly"dolly forward", "tracking shot"Follows subject movement
Drone"aerial view", "bird's eye pull back"Epic scale and perspective
Static"static camera", "fixed position"Focuses on subject motion
Handheld"handheld camera", "slight camera shake"Documentary/realistic feel

Step 4: Lighting & Atmosphere

Lighting keywords tell the AI what time of day and what atmospheric conditions to maintain or enhance:

  • Golden hour: warm, golden sunlight, long shadows, orange sky glow
  • Night: moonlight, star reflections, city lights, neon glow
  • Storm: dramatic clouds, lightning flashes, rain streaks, dark atmosphere
  • Fog: misty atmosphere, soft light diffusion, reduced visibility depth
  • Studio: clean lighting, soft shadows, professional photography style

Step 5: Style & Quality Keywords

Always end your prompt with style and quality modifiers:

  • Quality: "photorealistic", "ultra-detailed", "8K quality", "sharp focus"
  • Film style: "cinematic", "film grain", "anamorphic lens", "35mm film look"
  • Speed: "slow motion", "real-time", "time-lapse effect"
  • Rendering: "ray-traced lighting", "HDR", "volumetric lighting"

Real Prompt Examples

Example 1: Nature Scene

🌅 Mountain Lake
A serene mountain lake surrounded by pine trees, the water surface gently rippling with small waves, a light breeze causing the trees to sway softly, morning mist slowly lifting from the water, slow zoom out camera movement revealing the full mountain landscape, golden morning sunlight, cinematic 4K, photorealistic, ultra-detailed

Example 2: Portrait

👤 Portrait Animation
A young woman with curly auburn hair, her eyes slowly blinking and a gentle smile forming on her lips, hair strands delicately moving in a soft breeze, static camera, warm afternoon sunlight from the left side, shallow depth of field with creamy bokeh background, photorealistic, natural skin texture, 4K portrait

Example 3: Urban Scene

🌆 City at Night
A busy city intersection at night, cars moving through frame leaving light trails, pedestrians walking quickly, neon signs pulsing and reflecting on wet pavement, slow dolly forward camera movement, rain falling softly, dramatic cinematic night photography, ultra-sharp, 4K HDR, anamorphic lens distortion

Common Mistakes to Avoid

  1. Too vague: "Make it move" — specify exactly what moves and how
  2. Conflicting instructions: "Static camera that pans left" — camera can't be both
  3. Over-stuffing: 50+ words often causes the AI to ignore most instructions — aim for 30-60 words
  4. No motion direction: Say where something moves to, not just that it moves
  5. Forgetting atmosphere: Without lighting/atmosphere keywords, outputs look flat

Tool-Specific Tips

Runway Gen-3: Responds well to "camera: [movement]" syntax. Supports very precise camera control with words like "orbit," "crane up," and "arc shot."

Pika Labs: Shorter prompts work better (under 40 words). Use their Pikaffects for special physics effects.

Kling AI: Excellent with natural motion descriptions. Very good at understanding human body language prompts.

Stable Video Diffusion: Keep prompts focused on a single main motion. Handles environment motion better than character motion.

🎓 Practice Exercise: Take any photo on your phone and write a prompt using the formula above. Generate the video, note what worked and what didn't, then refine. After 5 iterations with the same image, you'll develop strong intuition for what the AI responds to.