88Z AI, 16 April — Tencent has officially released and open-sourced the Hunyuan 3D World Model 2.0. It supports multi-modal inputs to generate complete 3D worlds, exports assets in various formats, and integrates seamlessly with major game engines. The new version uses the WorldMirror 2.0 architecture, significantly improving visual realism and supporting free character exploration and physical collision.
At a glance
- Supports multi-modal input to generate editable 3D assets, seamlessly integrating with Unity/UE engines
- Uses WorldMirror 2.0 architecture to improve image quality and support physical collision and free exploration
The update
Tencent has officially launched and open-sourced the Hunyuan 3D World Model 2.0 (HY-World 2.0). This model features robust multi-modal processing capabilities, accepting text, images, and video as inputs to one-click generate, reconstruct, and simulate complete 3D worlds. The generated assets can be exported in editable formats such as Mesh, 3DGS, or point clouds, enabling seamless integration with mainstream game development tools like Unity and Unreal Engine. Compared to version 1.0, version 2.0 introduces the new WorldMirror 2.0 architecture, achieving a significant leap in image detail and realism. Furthermore, the new model adds a character mode that allows users to explore freely within the generated worlds and supports realistic physical collision effects, further enhancing interactive experience and application value.
Sources & editorial note
This article is an AI-assisted synthesis by 88Z AI based on the public sources below. Product claims and availability are attributed to their publishers. It does not represent independent testing or first-hand reporting.
- 腾讯混元Source published
Source dates above refer to the original announcements; the date under the headline is this article’s publication on 88Z AI.