Seedance 2.0 is a comprehensive production pipeline for quad-modal AI filmmaking, enabling users to generate and direct cinematic videos from text, images, or existing video. It targets AI filmmakers and agent builders, offering a modular, skill-based architecture that provides fine-grained control over cinematic elements like camera work, motion, lighting, and character fidelity, while optimizing AI agent efficiency.

How It Works

The core of Seedance 2.0 is its modular "skill constellation" architecture, comprising 23 independently loadable sub-skills. This design allows AI agents to precisely select and load only the necessary modules for a given task, minimizing token waste and maximizing precision. It offers two distinct workflows: "Max Detail" for professional, complex scenes requiring maximum creative control, and "Max Performance" for rapid iteration and short-form content, which currently favors short Chinese prompts for optimal results.

Quick Start & Requirements

Installation is streamlined via various AI CLI tools, including Antigravity, Gemini CLI, Claude Code, GitHub Copilot/Codex, Cursor, Windsurf, and OpenCode, using the command skills install https://github.com/Emily2040/seedance-2.0. Manual installation paths are detailed for multiple platforms and tools. No specific hardware (e.g., GPU, CUDA) or software prerequisites beyond the CLI tools are explicitly listed in the README.

Highlighted Details

• Quad-Modal Generation: Supports Text-to-Video, Image-to-Video, Video-to-Video, and Reference-to-Video.
• Modular Skill Architecture: 23 independently loadable skills for precise control over cinematic elements.
• Dual Workflows: "Max Detail" for professional, long-form production and "Max Performance" for quick, short-form iteration.
• Multilingual Support: Dedicated cinematic vocabulary modules for Chinese (400+ terms), Japanese, Korean, Spanish, and Russian.
• Advanced Cognitive Skills: Features include seedance-interview (L8) for constructive drafting and seedance-troubleshoot (L11) for identifying fundamental trade-offs.

Maintenance & Community

The project is authored and maintained by Emily (@iamemily2050), with contributions welcome via pull requests. Links to the author's social profiles (𝕏, IG) are provided. Version history is detailed in CHANGELOG.md. No explicit community forums (e.g., Discord, Slack) or roadmap links are present.

Licensing & Compatibility

Seedance 2.0 is released under the MIT license, which is permissive and generally compatible with commercial use and linking within closed-source projects.

Limitations & Caveats

The global API release was delayed due to copyright enforcement actions, and real-person face uploads were paused as of February 15, 2026. Content filters have been tightened, requiring the use of seedance-copyright before generation. The system currently performs significantly better with short prompts (under 2000 characters) in Chinese, indicating potential limitations for other languages or longer prompt structures.