Seedance 2.0 Prompt Engineering Advanced Guide
Seedance 2.0 Prompt Engineering Advanced Guide
1. Defining Subjects
Why Define Subjects
In reference materials, a single image often contains multiple subjects. To precisely reference a specific object in the material, clear subject definitions are necessary. Subjects can be people, props, scenes, etc.
Recommended Template
Define [subject core features] in <image/video N> as <subject N>
Core feature requirements: Use 2-3 clear, stable static features (such as clothing, hairstyle, appearance, category) to describe, ensuring unique identifiability.
Examples
Define the woman in the red dress and straw hat in image 1 as subject 1
Define the woman in the red dress and straw hat in image 1 as Zhang Hong
Usage Rules
#### Rule 1: Explicitly reference subjects each time they are involved
- Simple scenarios without predefined subjects: Use
to emphasize the binding relationship@
- Example: Zhang San@image 1
- Scenarios with predefined subjects: Use a consistent label to refer to them throughout
- Example: Define the tall man in video 1 as policeman, define the other short man as thief - Subsequently, consistently refer to the tall man as "policeman" and the short man as "thief"
#### Rule 2: Still use when using Asset IDs
The model cannot directly associate Asset IDs with reference content, so Asset IDs must not be used as a substitute.
#### Rule 3: Keep descriptions concise
Avoid redundancy and semantic conflicts (e.g., contradictory features for the same subject). For spatial relationships, it is recommended to express them through reference images first, reducing complex text descriptions.
2. Shot Sequencing
Core Concept
Seedance 2.0's internal modeling decouples space and time. Therefore, the ideal form for complex video prompts is timeline-based shot sequencing: break the video into several shots, dynamically describing each shot in chronological order of events.
Shot Formula
Who + Where + Doing What + How the Camera Moves
Practical Advice
Use shot ordering — write a simple "Shot 1 / Shot 2 / Shot 3" breakdown for each segment of the video, then merge them into a complete prompt.
Positive vs. Negative Examples
Negative example:
A man runs nervously on the street, the scene looks very cinematic.
Positive example:
Shot 1: Side shot in an alley, the man starts running slowly, with an urgent sense of breathing.
Shot 2: The man knocks over a fruit stand, the camera shakes quickly and cuts to a close-up of the man's panicked expression.
Shot 3: The man climbs over a low wall and disappears, the camera slowly pulls back and holds on the empty street.
Specific Rules
- Use
Shot 1,Shot 2,Shot 3etc., organizing content in chronological order (main action first, secondary later) - Do not strictly enforce duration for each segment — allow the model to naturally generate rhythm based on the plot
- The model's support for precise timing (e.g., 0-3 seconds) is unstable; forcing duration limits may lead to abnormal results
Organizational Logic for Each Shot
| Element | Description | Example |
| Camera movement or transition | How the shot begins | Slow full push-in, fixed camera position, cut to... |
| Subject action and expression | Key actions of the main character | Describe action, expression changes |
| Position or spatial change | Scene or spatial relationship | Where the subject is, positional changes |
| Audio information | Sound content | Sound effects, dialogue, or background music for the shot |
3. Action Description Requirements
Three Principles
#### Principle 1: Limb-level detail + Degree quantification
Actions must be specified down to hands, legs, head, shoulders, back, etc., supplemented by descriptions of range, speed, and intensity.
Examples: slowly raise hand, quickly turn head, forcefully push off ground, slightly lower head
#### Principle 2: Prioritize gentle, continuous, small movements
Prefer slow, soft, continuous subtle movements. Try to avoid high-intensity, large-motion actions like sprinting, big jumps, or violent rolling.
Examples: walk slowly, gently raise hand, slightly lower head, naturally sit down
#### Principle 3: Describe action transitions and connections
Specify the momentum and transition relationship between consecutive actions to ensure smooth and natural motion on screen.
Examples: using the momentum of turning around to naturally raise the hand, transitioning naturally from a paused state to raising the hand
Concrete Externalization of Emotions
Use specific physical details to express emotions, replacing abstract words like "very sad" or "extremely angry."
| Abstract Emotion | Externalized as Actions and Details |
| Sadness | Head lowered, shoulders trembling slightly, eyes reddening, fingers unconsciously gripping the hem of clothes, tears welling up but not falling |
| Joy | Corners of the mouth uncontrollably lifting, brows and eyes relaxing, steps becoming light, unconsciously humming a tune, unable to resist spinning around on the spot |
| Tension/Anxiety | Frequently checking the watch, fingers tapping the table incessantly, rapid breathing, eyes darting away, unconsciously biting fingernails |
| Anger | Fists clenched tightly, jawline tense, chest heaving violently, eyes sharp as knives, words squeezed out through gritted teeth |
| Relief | Letting out a long sigh, tense shoulders completely relaxing, a faint, long-lost smile appearing on the face, looking up into the distance |
4. Camera Movement Writing
Overview
The model has a strong understanding of camera movement terminology — standard terms can be used directly.
Common Camera Movement Terms
| Type | Terms |
| Shot size | Medium shot, close-up, full shot, medium close-up, long shot |
| Push/pull | Slow push-in, quick push-in, pull out, slow pull out |
| Pan/track | Smooth dolly, pan shot, follow shot |
| Fixed | Static shot, fixed camera position |
| Other | High angle, low angle, first-person view, over-the-shoulder shot |
Notes
Try to specify only 1 type of camera movement per shot. Requesting push, pull, pan and track simultaneously increases instability in the image.
Recommended Combinations
- Camera movement + subject action:
Medium shot smoothly tracks the girl as she walks briskly to the door - Camera movement + emotion:
Camera slowly pushes in for a close-up, expression slightly tense
🍅 Try AI Video Generation Free on Tomato AI
Sign up for free credits. Access Seeddance 2.0, Sora 2, Kling 3 & more top models. No watermark, 1080P output.
Start Creating Free →