'HY-World 2.0' has been released, which can generate 3D worlds for games, allowing users to create 3DCG worlds instead of videos and output them to Unity or UE5.
Tencent, a major Chinese technology company, has open-sourced 'HY-World 2.0,' a multimodal world model that can generate, reconstruct, and simulate interactive 3D worlds from text, images, and videos, as part of its '
HY-World 2.0: A Multi-Modal World Model for Reconstructing, Generating, and Simulating 3D Worlds
(PDF file) https://3d-models.hunyuan.tencent.com/world/world2_0/HY_World_2_0.pdf
HY-World 2.0 is a multimodal world model that can generate, reconstruct, and simulate interactive 3D worlds from text, images, and videos. Its output can be integrated into game engines and embedded simulation pipelines.
With a single click, you can automatically convert text and images into interactive 3D worlds. It can also export editable 3D worlds for game engines such as Unity and Unreal Engine. The exported 3D worlds include standard 3D export options (mesh, 3DGS, point cloud). An interactive character mode is supported, allowing you to explore the generated 3D world in real time.
We're open-sourcing HY-World 2.0, a multimodal world model that generates, reconstructs, and simulates interactive *3D worlds* from text, images, and videos.
— Tencent HY (@TencentHunyuan) April 16, 2026
Outputs can be integrated into game engines and embodied simulation pipelines.
Key highlights:
🔹 One-click world… pic.twitter.com/OuKEm9krn4
The 3D world generated by inputting the image in the lower right corner of the screen is shown below.
The following is the 3D world generated by the prompt 'Generate a retro voxel-style room with a fireplace'.
Input three images to generate a bedroom.
It's also possible to change the atmosphere of the generated 3D world with the press of a button.
HY-World 2.0 includes 'HY-Pano 2.0,' which scales up the generation of panoramas for high-fidelity 3D worlds from a single image; 'WorldNav,' which enables consistent exploration while avoiding collisions by performing semantic understanding-based trajectory planning using VLM and navigation meshes ; 'WorldStereo 2.0,' which enables stable generation of new viewpoints while maintaining spatially consistent memory through keyframe-based world expansion in latent space; 'WorldMirror 2.0,' an integrated 3D reconstruction method that can generate accurate and navigable 3DGS assets by integrating predictions from multiple viewpoints; and 'WorldLens,' a high-performance engine-independent 3DGS renderer for interactive exploration with lighting and collision handling capabilities.
Technical highlights from HY-World 2.0 👇
— Tencent HY (@TencentHunyuan) April 16, 2026
- 3D-first world modeling: a unified framework for world generation and reconstruction, built around spatial understanding in 3D.
- HY-Pano 2.0: scales panorama generation for high-fidelity 360° world initialization from single images —… https://t.co/ztAOEofVpy pic.twitter.com/VO6zXQLAUY
HY-World 2.0 is available for download from Hugging Face.
tencent/HY-World-2.0 · Hugging Face
https://huggingface.co/tencent/HY-World-2.0
It's also available on GitHub.
GitHub - Tencent-Hunyuan/HY-World-2.0: HY-World 2.0: A Multi-Modal World Model for Reconstructing, Generating, and Simulating 3D Worlds · GitHub
https://github.com/Tencent-Hunyuan/HY-World-2.0
Related Posts:
