Seedance 2 Prompt Engineering Skill

What This Skill Does

The Seedance 2 Prompt Engineering Skill, developed by dandysuper, is a powerful tool designed to help users create high-control, production-ready English prompts for the Seedance 2.0 and Seedance 2.0 Fast AI models. It specializes in leveraging multimodal references, including images, videos, and audio, to guide the AI's generation process. The skill assists in transforming rough ideas into detailed prompts, offering choices for input modes (Text-only, First/Last Frame, All-Reference). It meticulously maps assets (@asset) to specific controls, plans for 4-15 second durations with timeline beats, and supports multi-segment stitching for longer videos. Furthermore, it enables video extension, character replacement, directed editing, and accurate replication of camera language and motion rhythms from reference materials. Scenario-specific strategies are provided for diverse applications like product ads, short dramas, fantasy, and music videos.

Installation

To install this skill, use the following command in your OpenClaw environment:

clawhub install openclaw/skills/skills/dandysuper/seedance-2-prompt-engineering-skill

Use Cases

This skill is ideal for a wide range of users, including:

• Prompt Engineers: Refining rough ideas into precise, controllable prompts for Seedance 2.0.
• Content Creators: Generating unique video content for social media, marketing, or personal projects.
• Video Editors: Extending existing video clips, replacing characters, or directing scene edits.
• AI Artists: Exploring advanced prompt engineering techniques for multimodal AI generation.
• Developers: Integrating sophisticated video generation capabilities into applications.

Specific use cases include creating:

• Product advertisements with dynamic visuals.
• Short narrative dramas with character consistency.
• Immersive fantasy or sci-fi scenes.
• Visually compelling music videos.
• Stylized animated sequences based on reference art.

Example Prompts

Here are three examples of prompts a user might send to OpenClaw using this skill:

1. Create a prompt for Seedance 2.0 using @image1 (a cyberpunk city street at night) and @video1 (a slow pan across a bustling market). Mode: All-Reference. Focus on replicating the neon lighting and gritty atmosphere. Duration: 10s, Aspect Ratio: 16:9. Plan for a reveal of a character at 7s.
2. Generate a prompt for Seedance 2.0 Fast to extend a product shot of a smartwatch. Use @image1 as the product reference and @video1 as the desired motion style (smooth, professional product demo). Mode: First/Last Frame. Emphasize clean product presentation. Duration: 7s, Aspect Ratio: 1:1. Negative constraints: no distracting backgrounds.
3. Design a prompt for a short drama scene. Use @audio1 (dramatic orchestral music) for pacing and @image1 (a portrait of a determined warrior) for character identity. Mode: Text-only with Audio Pacing. Story: A warrior preparing for battle, followed by a charge. Duration: 12s, Aspect Ratio: 9:16. Include timecoded beats for action sequences.

Tips & Limitations

Tips:

• Be Specific: The more detailed your description of actions, camera movements, and visual styles, the better the AI will understand your intent.
• Utilize @asset: Clearly define what each reference asset controls (e.g., identity, style, motion).
• Structure is Key: Follow the recommended output format (Mode, Assets Mapping, Final Prompt, Negative Constraints, Generation Settings) for optimal results.
• Iterate: Start with a clear idea and refine the prompt based on the generated output.
• Understand Platform Limits: Be mindful of file type, number, and size restrictions, as well as total generation duration.

Limitations:

• File Limits: Seedance 2.0 has strict limits on the number and size of input files (images, videos, audio).
• Duration: Generation duration is capped between 4-15 seconds.
• Content Moderation: Prompts that violate copyright, feature recognizable characters/brands, or contain sensitive content may be rejected by the platform.
• Realistic Faces: Platform compliance might block generation using realistic human face references.
• Complexity: While the skill aids in complex prompting, achieving highly nuanced or abstract results may still require significant iteration and prompt refinement.