Seedance Prompt Skill

The skill supports Seedance 2.0's @ reference syntax to combine different input modalities in your prompts. This allows Claude to incorporate visual or auditory context into the generated text.

Supported Modalities

• Images (@图片1 to @图片9): Used for character reference, scene backgrounds, or first/last frame control.
• Videos (@视频1 to @视频3): Used for camera movement, action, special effects, or rhythm reference.
• Audio (@音频1 to @音频3): Used for background music, voice timbre, or sound effect reference.

Example Usage

If you have a product image and want an ad:

Input: 我有一张可乐产品图，想做一个炫酷的产品展示广告
Output: @图片1中的可口可乐饮料，360度高速旋转2圈后... (followed by prompt details)

The Seedance Prompt Skill is implemented as a single-file skill for Claude Code. The core logic and definitions reside in .claude/skills/seedance/SKILL.md. When installing, ensure this file is placed within the .claude/skills/seedance/ directory of your target project.

.
├── .claude/
│   └── skills/
│       └── seedance/
│           └── SKILL.md           # The skill definition (core file)
├── .gitignore
└── README.md

The skill is designed to be intuitive and can be activated in three ways:

1. Explicit Command: Use the /seedance command.
2. Keyword Detection: Mention keywords such as Seedance, 即梦, 视频提示词, 视频生成, AI视频, 短剧, 广告视频, or 视频延长.
3. Natural Language: Simply ask to "generate video prompts" or "create Seedance prompts".

Since Seedance 2.0 generates videos in 4-15 second increments, the skill uses a multi-segment strategy for longer content. It generates a primary segment and then provides a plan for 'Video Extension' using the previous segment as a reference.

Workflow for long videos:

1. Segment 1: Generate the initial 15-second clip.
2. Segment 2 (Extension): Upload the first clip as @视频1 and use the provided extension prompt to ensure continuity in action and setting.

To get professional results, use these advanced techniques supported by the skill:

• Timestamped Storyboarding: For 15-second videos, control specific segments by second (e.g., 0-3s: ..., 4-8s: ...).
• Technical Parameter Specification: Declare aspect ratio, frame rate, color tone, and camera style at the beginning of the prompt.
• Negative Prompting (Prohibition): Explicitly state elements to exclude (e.g., "no watermarks, no subtitles, no logos").
• Long Video Segmentation: For videos longer than 15 seconds, the skill automatically splits the request into multiple segments with "銜接點" (connection point) descriptions to allow for smooth extension via the video extension feature.
• Dialogue and Sound Effects: Write dialogue in quotes and include character emotions (e.g., 台词1 (角色名, 情绪): "内容").

If the generated prompts are not meeting your expectations, follow these strategies:

• Provide more specific descriptions and avoid abstract language.
• Use the Timestamp Storyboard Method (时间戳分镜法) to precisely control each visual segment.
• Add negative prompts (prohibited items) at the end of the prompt to exclude unwanted elements.
• Ask Claude to generate multiple versions so you can select and fine-tune the best one.

You can contribute by adding scenario-specific prompt examples, translating skill definitions for English support, adding new templates for platform updates, or optimizing prompt quality based on generation results.

Contribution Workflow:

1. Fork the repository.
2. Create a feature branch: git checkout -b feature/your-feature.
3. Commit your changes: git commit -m "Add your feature".
4. Push the branch: git push origin feature/your-feature.
5. Open a Pull Request.