A fully open-source multimodal canvas agent that implements Canva/Manus-style image and video creation in a local environment
Provides Magic Canvas / Magic Video, where the AI instantly understands intuitive actions like drawing, arrows, and doodles and generates results without a prompt
Handles image and video generation with a single prompt, supporting various models such as GPT-4o, Midjourney, VEO3, and Kling
Local models such as ComfyUI and Ollama can be used at the same time
Video/image editing and generation integrated into a single workflow
Provides infinite-canvas-based storyboard design, supporting scene linking, media management, and real-time collaboration
Supports conversational creation through an AI Agent system: object insertion, style switching, and logic control. Includes multi-character consistency
Flexible deployment structure supporting fully offline local or hybrid setups: Ollama + API integration possible
Provides built-in media and prompt libraries
Supports both Windows and macOS
Uses Python 3.12 or later, a React-based frontend, and the Vite build system
Provides development extensions for VSCode/Cursor environments, with npm- and pip-based installation