AI video generation tool that creates videos from text prompts, images, or reference clips with strong character consistency.
Matrix Game 2.0

About Matrix Game 2.0
Matrix Game 2.0 is an open-source AI foundation model designed to generate interactive real-time video in virtual environments at approximately 25 frames per second. The tool enables users to control characters and camera movements using simple keyboard commands, producing coherent minute-long video sequences across diverse visual styles and scenarios. It leverages advanced AI techniques to maintain physical consistency in character movements while adapting to various environments. Matrix-Game is particularly valuable for game developers prototyping virtual worlds, researchers exploring embodied intelligence, and content creators seeking rapid visualization tools. The AI translates intuitive user commands into immediate, accurate video generation with high controllability, temporal stability, and action responsiveness. Compared to previous models, it bridges the gap between content generation and interactive experiences by enabling real-time user interaction within generated video environments. The tool is open-source and typically requires technical knowledge to set up and run, as it is provided as a GitHub repository.
GitHub, Inc.
San Francisco, California, US · Founded 2008
- Founders
- Tom Preston-Werner, Chris Wanstrath, PJ Hyett, Scott Chacon
- Founded
- 2008
- Headquarters
- San Francisco, California, US
- Legal status
- Subsidiary of Microsoft (NASDAQ: MSFT)
Key features
- Generates interactive real-time video in virtual environments
- Supports keyboard-controlled characters and cameras
- Maintains physical consistency in character movements
- Adapts to diverse visual styles and scenarios
- Produces coherent minute-long video sequences
- Operates at approximately 25 frames per second
- Enables real-time user interaction within generated video
- Open-source foundation model
- High controllability and temporal stability
- Translates intuitive user commands into accurate video generation
Use cases
- Prototyping virtual worlds for game development
- Research in embodied intelligence and AI-driven environments
- Rapid visualization and content creation for diverse scenarios
Pros
- Generates real-time interactive video at approximately 25 frames per second
- Supports fine-grained control via keyboard and mouse inputs for dynamic scene evolution
- Provides pretrained model weights for universal scenes, GTA driving scenes, and TempleRun game scenes
- Leverages an auto-regressive diffusion-based image-to-world framework for coherent long video sequences
- Open-source and available on GitHub for community contributions and customization
Cons
- Requires technical knowledge to set up and run, including Linux OS and Nvidia GPU with at least 24GB memory
- Upward camera movement may cause brief rendering glitches in the current version
Frequently asked questions about Matrix Game 2.0
What is Matrix-Game 2.0?
Matrix-Game 2.0 is an open-source interactive world foundation model designed for real-time long video generation. It generates videos conditioned on keyboard and mouse inputs, enabling fine-grained control and dynamic scene evolution.
Who is Matrix-Game 2.0 suitable for?
The tool is suitable for game developers prototyping virtual worlds, researchers exploring embodied intelligence, and content creators seeking rapid visualization tools. It bridges the gap between content generation and interactive experiences.
What are the system requirements for Matrix-Game 2.0?
Matrix-Game 2.0 requires a Linux operating system, an Nvidia GPU with at least 24GB memory (tested on A100 and H100), and 64GB RAM. It also depends on dependencies like FlashAttention.
How do I get started with Matrix-Game 2.0?
To get started, clone the GitHub repository, create a Conda environment, install dependencies, download pretrained model weights, and run inference scripts like inference.py or inference_streaming.py with appropriate configurations.
Does Matrix-Game 2.0 support streaming video generation?
Yes, Matrix-Game 2.0 supports streaming video generation through the inference_streaming.py script, allowing users to generate interactive videos with their own input actions and images.
Are there any known issues with Matrix-Game 2.0?
Yes, upward camera movement may cause brief rendering glitches, such as black screens. A fix is planned for future updates.