This project offers an AI-driven workflow for generating multi-episode video series from novels or stories. It targets users aiming to leverage AI for short drama production, automating scriptwriting, asset description generation, image creation, storyboarding, and video assembly using Claude Code, Nana Banana Pro, and Seedance 2.0. The primary benefit is a streamlined, automated pipeline from text to finished video episodes.

How It Works

The core approach integrates Claude Code for script and storyboard generation, Nana Banana Pro (or other AI image generators) for visual assets, and Seedance 2.0 for video creation and extension. The workflow involves ideation, scriptwriting (four-act structure), asset planning, image generation, storyboard creation in Seedance 2.0's timeline format, and finally, video generation with seamless episode transitions using Seedance 2.0's video extension feature. This method aims to efficiently utilize AI capabilities for automated video production.

Quick Start & Requirements

• Primary install / run command: The project is designed to be invoked as a custom skill within an AI environment, e.g., ❯ seedance <story_name>. Direct installation instructions (pip, Docker) are not provided.
• Non-default prerequisites and dependencies: Requires integration with Claude Code and its skill system, along with the Seedance 2.0 platform. Nana Banana Pro or other image generation AI is optional. Hardware requirements for AI model execution (e.g., GPU) are implied but not specified.
• Links: Documentation and example projects are referenced via local paths within the repository (e.g., docs/, 林冲项目/). No external URLs for quick-start guides or demos are provided.

Highlighted Details

• Structured Workflow: Employs a clear pipeline: Theme -> Script -> Asset Descriptions -> Image Generation -> Storyboard -> Video Generation -> Video Extension.
• Asset & Script Naming: Utilizes strict file naming conventions for scripts (_剧本.md, _素材清单.md, _E[集数]_分镜.md) and a clear numbering system for assets (C## for characters, S## for scenes, P## for props).
• Seedance 2.0 Prompting: Features a specific prompt format including material slots, a timeline-based Seedance Prompt with duration, visual/audio cues, and references, plus a tail frame description for continuity.
• Seamless Episode Transitions: Achieved by using the previous episode's video as input (@视频1) in the prompt for the next episode's generation, enabling smooth continuity.
• Camera Control: Provides a list of keywords for various camera movements and effects (e.g., "推镜头" for push-in, "环绕镜头" for orbiting).

Maintenance & Community

The project welcomes feedback and suggestions for improvement. No specific contributors, sponsorships, community channels (Discord/Slack), or roadmaps are detailed.

Licensing & Compatibility

• License type: The project content is explicitly stated as "for learning and reference only" (仅供学习和参考使用). This is not a standard open-source license and implies significant restrictions, likely prohibiting commercial use or integration into proprietary systems.
• Compatibility notes: Due to the restrictive licensing, compatibility for commercial use or closed-source linking is highly questionable.

Limitations & Caveats

• Sensitive Word Filtering: The Seedance platform has an undisclosed sensitive word filter; users must manually identify and avoid triggering terms.
• Generation Limits: Seedance 2.0 generation is limited to a maximum of 9 images and 3 videos (totaling 15 seconds) per run.
• Prompt Complexity: Complex prompts exceeding 300 words may result in inconsistent instruction following.
• Video Editing: Video editing capabilities are limited, often requiring regeneration for modifications.
• Platform Dependency: The workflow is tightly coupled with Claude Code and the Seedance 2.0 platform, suggesting limited portability.