--- license: mit tags: - world model - video generation - game intelligence --- # MineWorld A Real-time interactive world model on Minecraft ## 🔧 Setup 1. Clone this repository and navigate to MineWorld folder: ```bash git clone https://github.com/microsoft/mineworld cd mineworld ``` 2. We provide an `requirements.txt` file for setting up a pip environment. ```bash # 1. Prepare conda environment conda create -n mineworld python=3.10 # 2. Activate the environment conda activate mineworld # 3. install our environment pip3 install -r requirements.txt ``` ## Inference Please refer to the [paper](https://arxiv.org/pdf/2504.08388) and [code](https://github.com/microsoft/MineWorld) for detailed inference instructions. ## 💡 Intended Uses Our model is solely trained in the Minecraft game domain. As a world model, an initial image in the game scene will be provided, and the users should select an action from the action list. Then the model will generate the next scene that takes place the selected action. ## 🪧 Out-of-scope Uses Our models are not specifically designed for any tasks or scenarios other than the Minecraft model. Developers should expect failures in generation results regarding the out-of-scope scenarios. Developers should be aware of and adhere to applicable laws or regulations (including privacy, trade compliance laws, etc.) that are relevant to their use case, and evaluate and mitigate for privacy, safety, and fairness before using within a specific downstream use case, particularly for high-risk scenarios. ## 🤖️ Risks and Limitations Some of the limitations of this model to be aware of include: * Quality of Service: MineWorld is trained solely on Minecraft, so it cannot generate results for other video domains (such as internet video). And the model cannot generate videos with higher resolution. * Information Reliability: MineWorld is trained on videos with a fixed resolution, therefore the results may lose detailed information due to the low resolution. * MineWorld inherits any biases, errors, or omissions characteristic of its training data, which may be amplified by any AI-generated interpretations. * MineWorld was developed for research and experimental purposes. Further testing and validation are needed before considering its application in commercial or real-world scenarios. * The input of other images than Minecraft will result in incoherent imagery being created and should not be attempted. * Users are responsible for sourcing their datasets legally and ethically. This could include securing appropriate copy rights, ensuring consent for use of audio/images, and/or the anonymization of data prior to use in research. ## License The model is released under the [MIT license](https://github.com/microsoft/mineworld/blob/main/LICENSE). ## ☎️ Contact We welcome feedback and collaboration from our audience. If you have suggestions, questions, or observe unexpected/offensive behavior in our technology, please contact us through `junliangguo AT microsoft.com`. ## BibTeX If you find our project helpful to your research, please consider starring this repository🌟 and citing our paper. ```bibtex @article{guo2025mineworld, title={MineWorld: a Real-Time and Open-Source Interactive World Model on Minecraft}, author={Guo, Junliang and Ye, Yang and He, Tianyu and Wu, Haoyu and Jiang, Yushu and Pearce, Tim and Bian, Jiang} year={2025}, journal={arXiv preprint arXiv:2504.08388}, } ```