HunyuanWorld (Tencent)

Tencent Hunyuan's open-source 3D world-generation series that turns text or an image into an explorable 3D scene.

HunyuanWorld is an open-source world-generation model series from Tencent's Hunyuan team. HunyuanWorld 1.0, from July 2025, uses a 360° panorama as an intermediate representation, layering the scene semantically and reconstructing it into an exportable 3D mesh, generating an explorable 3D world from a single sentence or image; Voyager, from September 2025, is an RGB-D video diffusion model that generates geometrically consistent color and depth video along a camera path the user provides, yielding a 3D point cloud directly. Later releases followed: 1.1 (WorldMirror, reconstruction from video or multi-view images), 1.5 (WorldPlay, December 2025, a real-time interactive world model), and HY-World 2.0 (April 2026). For embodied AI, it's used mainly to mass-produce simulation scenes and assets — it leans toward “generating a 3D environment,” a different focus from a world model that predicts the consequences of a robot's actions.

ExampleGiven the prompt “a seaside town of wooden cabins,” HunyuanWorld 1.0 generates an explorable 360° 3D scene with an exportable mesh, ready to import into a game engine or simulator.

Also called

HunyuanWorld 1.0, HunyuanWorld-Voyager, HY-World, WorldPlay

Related

World Model · Interactive World Model · Generative Simulation · Marble (World Labs) · Genie 3 · Spatial Intelligence

Sources

HunyuanWorld 1.0 technical report (arXiv 2507.21809)
Tencent-Hunyuan/HunyuanWorld-1.0 (GitHub, news timeline)
Voyager: Long-Range and World-Consistent Video Diffusion (arXiv 2506.04225)

As of

2026-05