README.md

metadata

title: HY-WorldPlay
emoji: 🌍
colorFrom: blue
colorTo: green
sdk: gradio
python_version: 3.1
app_file: app.py
hardware: zero-gpu
license: apache-2.0
short_description: HY-World 1.5 - Interactive World Modeling

HY-WorldPlay (HunyuanWorld 1.5)

This Space demonstrates HY-WorldPlay: a streaming video diffusion model for real-time interactive world modeling.

Model Details

• Base Model: HunyuanVideo-1.5
• Architecture: Latent Video Diffusion with Dual Action Representation
• Capability: Generates long-horizon streaming video with geometric consistency.

Usage

1. Prompt: Enter a text description of the scene.
2. Image: (Optional) Upload a starting image for Image-to-Video generation.
3. Camera Path: Upload a JSON file defining the camera trajectory (Pose).
   • Note: You can find example pose files in the official repository or use the default provided in the demo if implemented.
4. Generate: Click generate to create the video.

Notes

• This Space uses ZeroGPU for inference.
• The first run might take longer to download the model weights (~30GB+).
• The model is running in Bidirectional mode by default for quality.

Citation

@article{worldplay2025,
    title={WorldPlay: Towards Long-Term Geometric Consistency for Real-Time Interactive World Model},
    author={Wenqiang Sun and Haiyu Zhang and Haoyuan Wang and Junta Wu and Zehan Wang and Zhenwei Wang and Yunhong Wang and Jun Zhang and Tengfei Wang and Chunchao Guo},
    year={2025},
    journal={arXiv preprint}
}