Skip to main content
Video prompts work differently from image prompts. The scene is already defined — by the first frame in Image-to-Video, or by the subject description in Text-to-Video. The prompt’s job is to describe movement, not appearance.

Prompt structure for video

Write elements in this order:
PositionElementExample
1Subject motionwalking slowly forward
2Secondary motionhair blowing in the wind, dress flowing
3Camera movementslow zoom in, steady camera
4Lighting or moodsoft natural light, warm tone
Keep it short. Four to eight terms produce more consistent results than long descriptive paragraphs.

Motion terms that work

Body motion: walking forward, turning head slowly, looking toward camera, sitting down, leaning back, raising arms, subtle breathing motion, blinking Secondary motion: hair blowing, dress flowing, water rippling, leaves moving, smoke drifting Camera: slow zoom in, slow zoom out, steady camera, slow pan left, slow pan right, slight camera shake, pull back, push in, orbit around subject Pace: slow motion, normal speed, gentle, subtle

Good vs bad video prompts

Bad: beautiful woman, soft light, photorealistic, high quality — this describes an image, not motion. The model produces a near-static clip. Good: slow pan right, subject turns head toward camera, hair moving gently, steady handheld camera feel, warm soft light Bad: running, jumping, spinning, dancing, waving, laughing — too many simultaneous actions. The model can’t resolve them and produces chaotic or distorted motion. Good: walking slowly forward, slight sway in movement, camera pulls back steadily

Image-to-Video vs Text-to-Video prompting

Image-to-Video — the scene is fixed. Focus entirely on motion and camera. Don’t redescribe the subject or setting. Text-to-Video — the scene is not fixed. Open with a brief subject and setting description, then move into motion and camera terms. See Text-to-Video.

Tips

  • One primary motion per prompt — describe one main action and let secondary motion support it
  • Add camera movement explicitly — without it the model defaults to a locked, static camera
  • Avoid conflicting directions — zoom in and pull back in the same prompt cancel each other out
  • Subtle motion prompts produce more stable output than large movement prompts — slow turn beats spinning around
  • If the output is too static, add continuous motion, fluid movement to the prompt

Image-to-Video

Animate a first frame using a motion prompt.

Text-to-Video

Generate a clip from prompt alone.

Prompting Basics

Image prompt structure for reference.