Seedance 2.0 Prompter

Seedance 2.0 Prompt Designer Skill

This skill transforms a user's scattered multimodal assets (images, videos, audio) and ambiguous creative intent into a structured, executable prompt for the Seedance 2.0 video generation model. It acts as an expert prompt engineer, ensuring the highest quality output from the underlying model.

When to Use

• When the user provides multiple assets (images, videos, audio) for video generation.
• When the user's request is complex and requires careful prompt construction.
• When using the Seedance 2.0 model for video generation.

Core Function

This skill analyzes all user inputs and generates a single, optimized JSON object containing the final prompt and recommended parameters. The internal workflow (Recognition, Mapping, Construction) is handled automatically and should not be exposed to the user.

Usage Example

User Request: "Make the Mona Lisa drink a Coke. I want it to feel cinematic, like a close-up shot." User uploads `monalisa.png` and `coke.png`

Agent using `seedance-2.0-prompter`: The agent internally processes the request and assets, then outputs the final JSON to the next skill in the chain.

Final Output (for internal use):

{
  "final_prompt": "A cinematic close-up shot of a woman picking up a bottle of Coke and taking a sip. The scene is lit with dramatic, high-contrast lighting. Use @monalisa as the subject reference, and the object appearing in the video is @coke.",
  "recommended_parameters": {
    "duration": 8,
    "aspect_ratio": "16:9"
  }
}

Knowledge Base

This skill relies on an internal knowledge base to make informed decisions. The agent MUST consult these files during execution.

• `/references/atomic_element_mapping.md`: Core Knowledge. Contains the "Asset Type -> Atomic Element" and "Atomic Element -> Optimal Reference Method" mapping tables.
• `/references/seedance_syntax_guide.md`: Seedance 2.0 "@asset_name" syntax reference.
• `/references/prompt_templates.md`: Advanced prompt templates for different genres (e.g., Cinematic, Product Showcase, Narrative).

Atomic Element Mapping Knowledge Base

This document provides the core logic for the Seedance Prompt Designer skill. It contains two critical mapping tables.

Table 1: Asset Type -> Potential Atomic Elements

This table maps the type of user-uploaded asset to the most likely atomic element roles it can play in video generation. The skill should use this table in Phase 1 to analyze uploaded assets.

Table 2: Atomic Element -> Optimal Reference Method

This table defines the best way to reference each atomic element when constructing the prompt. The skill must use this table in Phase 2 to design the reference strategy.

Prompt Templates for Seedance 2.0

These are high-level templates that can be adapted by the skill to create more sophisticated and stylized prompts.

Template 1: Cinematic Character Shot

Goal: Create a short, film-like shot focusing on a character.

Template:

A cinematic [Shot Type] of [Character Description]. The scene is lit with [Lighting Style] lighting, creating a [Mood/Atmosphere] mood. The camera performs a [Camera Movement].

@character_ref.png is the subject reference.
@style_ref.png is the aesthetic style reference.

Example Usage:

• Shot Type: close-up
• Character Description: a pensive android
• Lighting Style: dramatic, high-contrast
• Mood/Atmosphere: somber and introspective
• Camera Movement: slow push-in

Template 2: Dynamic Product Showcase

Goal: Create an engaging video to showcase a product.

Template:

A dynamic, high-energy shot of [Product Name]. The product performs [Product Action]. The background is a [Background Description]. The video features fast-paced cuts and [Visual Effect].

@product_image.png is the product reference.
@motion_ref.mp4 is the motion reference for the product action.
@music_track.mp3 provides the energetic soundtrack.

Example Usage:

• Product Name: the new "Quantum" smartphone
• Product Action: rotating 360 degrees while emitting light trails
• Background Description: clean, minimalist studio setting
• Visual Effect: lens flares

Template 3: Narrative Scene

Goal: Generate a short scene that tells a simple story.

Template:

[Scene Description]. [Character 1 Description] [Action 1]. Then, [Character 2 Description] [Action 2]. The scene should feel [Overall Mood].

@character1.png is the reference for the first character.
@character2.png is the reference for the second character.
@location.png is the scene reference.
@dialogue.mp3 provides the audio for the scene.

Example Usage:

• Scene Description: In a dimly lit library
• Character 1: a young wizard
• Action 1: nervously looks around before opening an ancient book
• Character 2: a ghostly figure
• Action 2: appears behind him
• Overall Mood: mysterious and suspenseful

Seedance 2.0 Syntax Guide

This document covers the essential syntax for interacting with the Seedance 2.0 model, focusing on the @ reference system.

The @ Reference System

The @ symbol is used to tell the model how to use the uploaded assets. It allows you to assign specific roles to your images, videos, and audio files.

Basic Syntax: @asset_name

• asset_name corresponds to the filename of the uploaded asset.

Common Usage Patterns

1. Defining Subject Identity:

• Prompt: A man wearing a hat. Use @man_photo.jpg as the subject reference.
• Explanation: This tells the model to use the person from man_photo.jpg as the main character.

2. Defining Scene/Environment:

• Prompt: A dog is running on the beach. Use @beach_scenery.png as the background.
• Explanation: This sets the scene from beach_scenery.png as the environment for the action.

3. Referencing Motion:

• Prompt: Make the character dance. Reference the motion from @dance_video.mp4.
• Explanation: The character in the generated video will mimic the dance moves from dance_video.mp4.

4. Referencing Camera Language:

• Prompt: A slow zoom-in on the product. Use the camera movement from @zoom_shot.mp4.
• Explanation: The generated video will replicate the camera motion of zoom_shot.mp4.

5. Using Start/End Frames:

• Prompt: An animation of a logo appearing. Use @logo_start.png as the start frame and @logo_end.png as the end frame.
• Explanation: The video will animate the transition from the start image to the end image.

6. Combining Multiple References:

• Prompt: Make the person from @person.jpg perform the action from @action.mp4 in the location from @location.png.
• Explanation: This is a powerful way to combine identity, motion, and scene from different sources.

7. Audio Reference:

• Prompt: A video of a dramatic landscape, with the music from @epic_music.mp3.
• Explanation: The model will use the provided audio as the soundtrack and may even try to match the video's pacing to the music's rhythm.

Best Practices

• Be Explicit: Clearly state the role of each @ reference. Don't assume the model will guess correctly. For example, instead of just ... @my_video.mp4, write ... using the camera motion from @my_video.mp4.
• One Role Per Asset (Usually): While an asset can sometimes serve multiple roles (e.g., a video providing both motion and a subject), it's clearer to assign one primary role per asset.
• Name Assets Clearly: Use descriptive filenames for uploaded assets (e.g., character_ref.png, camera_dolly_in.mp4). This makes the prompt easier to read and debug.
• Check for Conflicts: Ensure you are not giving contradictory instructions (e.g., referencing two different videos for the same motion).