SeedanceSeedance

Seedance Prompt Guide

Master the art of prompt writing for both image-to-video and text-to-video generation. Learn how to create compelling videos with precise control over motion, camera work, and storytelling.

Table of Contents

Image-to-Video Techniques

Text-to-Video Techniques

Key Summary

Essential formulas and techniques from the original guide

  • • Language recommendation: Prompts are recommended to use English or Chinese
  • • Negative prompts do not work with this model
  • • Camera language response is a strength of Seedance 1.0
  • • Be explicit about motion intensity - the model cannot infer this from static images
  • • Select "non-fixed camera" when using camera movement prompts
  • • All image-to-video techniques also apply to text-to-video

Image-to-Video Techniques

Transform your static images into dynamic videos with these proven prompt strategies

1. Basic Prompt Structure

Prompt = Subject + Motion, Background + Motion, Camera + Motion...

Since the scene already exists in your image, reduce (or even avoid) descriptions of static/unchanging parts. Focus on describing the moving parts, including subject motion, background motion/changes, and camera movement. The model will expand prompts based on your expression and understanding of the image to generate expected videos.

Key Principles:

✓ Essential Guidelines
  • Basic structure: Clearly describe the moving subject, background changes, and camera movements
  • Simple and direct: Use simple words and sentence structures
  • Feature description: Add distinctive features like "elderly man" or "woman wearing sunglasses" to better locate the subject
  • Key degree adverbs: Be explicit about intensity - "quickly", "dramatically"
✗ Common Mistakes
  • Contradicting the image: e.g., writing "a woman dancing" when the image shows a man
  • Mismatching backgrounds: e.g., writing "singing in a café" when the background is grassland
  • Adding non-existent elements: e.g., mentioning "the hand with accessories" when there are no accessories
  • Contradicting camera settings: e.g., writing camera movement when "fixed camera" is selected

2. Sequential Actions

The model has strong response to multi-shot actions, supporting temporal sequences of multiple consecutive actions and different actions by multiple subjects. You can try writing:

Prompt = Subject1 + Motion1 + Motion2
Prompt = Subject1 + Motion1 + Subject2 + Motion2...

List them in sequence, and the model will expand prompts based on your expression and understanding of the image to generate expected videos.

3. Camera Movement & Shot Transitions

You can use natural language in prompts to describe the camera changes you want, supporting orbit, aerial shots, zoom, pan, follow, handheld and other camera movements, as well as shot transitions. Camera language response is a strength of Seedance 1.0.

Camera Movements

  • Orbit: Camera circles around subject
  • Aerial: Bird's eye view shots
  • Zoom: Push in/pull out movements
  • Pan: Horizontal camera movement
  • Follow: Camera tracks the subject
  • Handheld: Natural camera shake

Shot Transitions

  • When writing consistent multi-shot prompts, write the internal connections between shots
  • Camera changes are connected through the clear prompt "shot transition"
  • If the scene changes after cutting, describe the new scene
  • When there are camera movement prompts, select "non-fixed camera" in basic parameters

Important Notes

  • Universal application: Camera movement prompts also work for text-to-video scenarios
  • Parameter selection: Choose "non-fixed camera" when using camera movement prompts
  • Scene description: Describe new scenes after shot transitions

4. Intensity Enhancement

If you want to highlight action frequency and intensity, or subject characteristics in the video, use degree adverbs appropriately.

Key Principles

  • Be explicit: The model cannot get motion intensity from the input reference image, so it must be clear in the prompt, otherwise the model will supplement according to its own understanding, which may deviate from user intent. e.g., "car drives by" → "car drives by quickly"
  • Amplify appropriately: You can appropriately exaggerate the degree to enhance video expressiveness. e.g., "man roars" → "man roars frantically", "wings flap" → "wings flap dramatically"

Degree Prompt Words

quicklydramaticallyvigorouslyhigh frequencypowerfullyfrantically

...and other intensity descriptors

Text-to-Video Techniques

Create videos from scratch using detailed text descriptions

1. Core Structure

Prompt = Subject + Motion + Scene + Camera & Style...

Subject + Motion + Scene are the most core and basic elements. The model will expand prompts to generate expected videos.

2. Advanced Description Techniques

Character Details

Focus on appearance, clothing, and posture for rich character development.

Example: "A young woman with flowing black hair, wearing a traditional red dress, standing gracefully with hands clasped"

Emotion & Dynamics

Combine character emotions with environmental dynamics for rich storytelling.

Example: "The warrior's determined expression as storm clouds gather overhead, wind whipping through his cloak"

Environment Details

Detailed descriptions of natural or architectural environments.

Example: "Ancient temple with carved stone pillars, flickering candlelight casting dancing shadows on weathered walls"

Atmospheric Lighting

Use lighting descriptions to enhance mood and visual appeal.

Examples: "golden hour", "dawn light", "dim atmosphere", "warm glow", "ethereal moonbeams"