This project provides a custom skill for Claude Code, transforming it into an AI video prompt engineer specifically for ByteDance's Seedance 2.0 platform. It enables users to generate structured, production-ready Chinese video prompts from natural language descriptions, streamlining the creation of cinematic AI videos for various applications like advertising, short dramas, and fantasy animations.

How It Works

The skill acts as an intermediary, interpreting user ideas and translating them into Seedance 2.0's specific prompt syntax. It leverages ten core prompt generation capabilities, including pure text generation, consistency control via reference images, camera and motion replication from reference videos, story completion, video extension, sound design, one-take long shots, video editing, and music beat synchronization. Advanced features like timestamp storyboarding for precise timing, technical parameter specification (aspect ratio, frame rate, color grading), negative prompting, and multi-segment stitching for videos exceeding 15 seconds are supported. It also integrates a multi-modal reference system, allowing users to incorporate images, videos, and audio into their prompts.

Quick Start & Requirements

• Prerequisites: Claude Code CLI installed and authenticated; a Seedance 2.0 account at jimeng.jianying.com.
• Installation: Clone the repository and copy the SKILL.md file to the .claude/skills/seedance/ directory within your project or globally to ~/.claude/skills/seedance/.
• Usage: Initiate Claude Code in your project directory and use the /seedance command or describe your video idea. The skill automatically activates with keywords like "Seedance" or "即梦".
• Links: Seedance 2.0 platform: jimeng.jianying.com

Highlighted Details

• Ten Core Capabilities: Covers a wide range of video generation needs from basic text prompts to complex editing and synchronization.
• Multi-Modal Reference System: Supports up to 9 images, 3 videos, and 3 audio files per prompt for rich context.
• Advanced Prompting: Features timestamp storyboarding, technical parameter specification, negative prompting, and multi-segment planning for videos over 15 seconds.
• Scenario-Specific Strategies: Tailored prompt generation approaches for e-commerce, short dramas, fantasy animation, science education, and music videos.
• Built-in Asset Libraries: Curated vocabularies for camera language and visual styles enhance prompt quality.

Maintenance & Community

Contributions are welcomed for adding prompt examples, translation, and supporting new Seedance features. Specific community channels or active maintainer information are not detailed in the README.

Licensing & Compatibility

• License: MIT.
• Compatibility: The MIT license permits broad use, including commercial applications. Users must adhere to the Seedance 2.0 platform's terms of service.

Limitations & Caveats

The Seedance 2.0 platform prohibits the upload of images or videos containing realistic human faces. Video generation is typically limited to 4-15 seconds per segment, requiring multi-segment planning for longer content.