Intermediate

Video Prompting Techniques

Learn to write effective prompts that control motion, camera movement, duration, and visual style to achieve cinematic AI-generated videos.

Anatomy of a Video Prompt

An effective AI video prompt has several key components that work together to guide the generation model. Unlike image prompts, video prompts must also describe motion and temporal changes.

💡
Prompt formula: [Camera movement] + [Subject description] + [Action/Motion] + [Environment] + [Lighting/Mood] + [Style]. Example: "Slow tracking shot of a woman walking through a neon-lit Tokyo alley at night, rain reflecting city lights on wet pavement, cinematic 35mm film look."

Motion Control

Controlling motion is critical for professional-looking AI videos. Use specific motion descriptors:

  • Subject motion: "walking slowly," "running toward camera," "turning head left," "gesturing with hands"
  • Environmental motion: "wind blowing through trees," "waves crashing," "clouds drifting," "traffic flowing"
  • Speed descriptors: "slow motion," "time-lapse," "real-time," "hyperspeed"
  • Motion quality: "smooth," "fluid," "jerky," "stop-motion style"

Camera Directions

Use cinematographic terminology to control camera behavior:

Camera TermDescriptionExample Usage
PanCamera rotates horizontally"Pan left revealing the cityscape"
TiltCamera rotates vertically"Tilt up from feet to face"
Dolly/TrackCamera moves forward/backward"Dolly in toward the subject"
Crane/BoomCamera rises or lowers"Crane shot rising above the forest"
OrbitCamera circles the subject"360-degree orbit around the statue"
HandheldSlight natural shake"Handheld documentary-style close-up"
StaticCamera stays fixed"Static wide shot, subject moves through frame"
Drone/AerialBird's eye perspective"Aerial drone shot flying over mountains"

Duration and Pacing

Different platforms support different clip lengths. Optimize your prompts accordingly:

  • Short clips (2-4s): Focus on a single action or moment. Best for loops and social media
  • Medium clips (5-10s): Allow for a sequence of movements or a scene to develop
  • Long clips (10-60s): Describe scene progression, but keep each "beat" simple

Style Keywords

Control the visual aesthetic with style keywords:

  • Cinematic: "cinematic," "35mm film," "anamorphic lens," "shallow depth of field," "bokeh"
  • Photorealistic: "photorealistic," "hyperrealistic," "8K," "RAW photo"
  • Stylized: "anime," "watercolor," "oil painting," "pixel art," "claymation"
  • Lighting: "golden hour," "neon-lit," "dramatic chiaroscuro," "soft diffused light," "volumetric fog"
  • Mood: "moody," "ethereal," "dystopian," "whimsical," "noir"

Prompt Examples

💡
Beginner prompt: "A golden retriever running on a beach at sunset."
✅
Advanced prompt: "Slow-motion tracking shot following a golden retriever sprinting along a pristine beach at golden hour, wet paws splashing through shallow surf, warm backlight creating a lens flare, shallow depth of field with soft bokeh background, cinematic 35mm film grain, 4K."

Common Mistakes to Avoid

  • Too vague: "A cool video" gives the AI no direction. Be specific about subject, action, and environment
  • Too complex: Cramming too many elements and actions into one clip leads to incoherent results
  • Ignoring physics: Requesting physically impossible actions can break visual coherence
  • Conflicting styles: Combining contradictory styles (e.g., "photorealistic anime") confuses the model
  • No motion cues: Forgetting to describe movement results in near-static video