MagicWorld: Interactive Geometry-driven Video World Exploration

     

We introduce MagicWorld, an interactive video world model that supports exploring a dynamic scene created from a single scene image through continuous keyboard actions, while maintaining structural and temporal consistency. MagicWorld generates action-driven point clouds from user inputs (W, A, S, D) to provide geometric constraints for stable viewpoint transitions.

📣 News

• 2026/02/10: We open-source the MagicWorld v1 codebase, including training and inference scripts.
• 2025/11/24: Our Paper on ArXiv is available 🥳!

✅ To-Do List for MagicTryOn Release

• ✅ Release the source code
• [ ] Release the MagicWorld v1 pretrained weights
• [ ] Release the MagicWorld v2

😍 Installation

Create a conda environment & Install requirments

# python==3.12.9 cuda==12.3 torch==2.2
conda create -n magicworld python==3.12.9
conda activate magicworld
pip install -r requirements.txt

If you encounter an error while installing Flash Attention, please manually download the installation package based on your Python version, CUDA version, and Torch version, and install it using pip install flash_attn-2.7.3+cu12torch2.2cxx11abiFALSE-cp312-cp312-linux_x86_64.whl.

😉 Demo Inference

Before inference, you need to do two things: (1) install the Uni3C library and its environment, then import the path to your installed Uni3C in uni3c_cam_render_api.py. (2) run action2traj.py to map your keyboard actions to a camera trajectory and generate the trajectory .txt file.

python inference/interactive_magicworld_v1.py

😘 Acknowledgement

Our code is modified based on VideoX-Fun. We adopt Wan2.1-I2V-14B as the base model. We use Uni3C to generate 3D points. Thanks to all the contributors!

😊 License

All the materials, including code, checkpoints, and demo, are made available under the Creative Commons BY-NC-SA 4.0 license. You are free to copy, redistribute, remix, transform, and build upon the project for non-commercial purposes, as long as you give appropriate credit and distribute your contributions under the same license.

🤩 Citation

@article{li2025magicworld,
  title={Magicworld: Interactive geometry-driven video world exploration},
  author={Li, Guangyuan and Zheng, Siming and Xu, Shuolin and Chen, Jinwei and Li, Bo and Hu, Xiaobin and Zhao, Lei and Jiang, Peng-Tao},
  journal={arXiv preprint arXiv:2511.18886},
  year={2025}
}