Yuuhann1999/agent-storyboard

What it solves

Creating short videos often involves fragmented workflows where scripts, storyboards, images, and voiceovers are scattered across different documents and tools. Agent Storyboard centralizes this process into a single, local-first storyboard table, allowing AI agents to handle the generation of assets while the user focuses on reviewing and adjusting the final output.

How it works

The project operates as a local web service that communicates with AI agents (like Codex or Claude Code) via the Model Context Protocol (MCP). The workflow follows these steps:

1. Project Creation: An agent creates a full storyboard project based on a user's theme and requirements.
2. Review: The user checks and modifies the storyboard table (including dialogue and visual descriptions).
3. Generation: The agent generates images, motion videos (via HyperFrames/Remotion), or voiceovers based on the storyboard entries.
4. Auto-fill: Generated assets are automatically linked back to the corresponding shots in the storyboard table.

Who it’s for

Content creators and video producers who want to automate the tedious parts of video pre-production and asset generation using AI agents without losing manual control or data privacy.

Highlights

• Local-first Architecture: All projects, scripts, and assets are stored locally on the user's machine.
• Agent-Driven Workflow: Agents can directly write projects and generate assets without requiring browser control.
• Integrated Voiceover: Supports voice cloning, style descriptions, and automatic dialogue alignment using local Whisper for timing.
• Coded Motion: Integrates with HyperFrames and Remotion for rendering subtitles and infographics via code.
• Visual Consistency: Includes a style library and a DESIGN.md file to ensure all generated assets follow a consistent visual specification.
• Flexible Layouts: Features a responsive storyboard table with compact and card views for different screen sizes.