Tencent HY-World-2.0 Turns Photos Into 3D Worlds

By vramkickedin | | 2 min read
A large translucent sphere constructed entirely of shimmering interlocking polygonal meshes with a realistic landscape inside the sphere.

Tencent has released HY-World-2.0, an open-source framework that transforms text, single photographs, and video clips into navigable three-dimensional spaces. Rather than generating short playback loops, the system builds complete digital models that creators can modify and load directly into standard design applications.

This release shifts focus from temporary media formats to stable scene files that process efficiently on local machines. Accepting varied visual inputs, the system gives independent teams a way to prototype environments without relying on cloud-based rendering services.

Model Size: 5GB & VRAM GPU: requirements vary

Core capabilities for local scene creation

  • Generates complete environments from simple text prompts or single photographs.
  • Reconstructs digital spaces by analyzing casual video footage.
  • Supports physics-based movement so users can explore scenes safely.
  • Outputs editable files that load directly into popular game engines.
  • Runs efficiently on a single graphics card with flexible resolution options.

Independent builders designing training simulations or architectural previews will benefit from the offline workflow. Processing images locally removes reliance on external servers while the streamlined pipeline reduces manual setup steps.

Project roadmap and technical limits

The team confirms that navigation and expansion modules arrive in future updates, leaving early access limited to reconstruction tasks. Users must currently combine external software for full scene generation until the complete toolkit launches.

"HY-World 2.0 takes a fundamentally different approach: it directly produces editable, persistent 3D assets (meshes / 3DGS) that can be imported into game engines like Blender/Unity/Unreal Engine/Isaac Sim,"

noted the team in a project post. This phased release allows early adopters to verify accuracy while developers finalize the remaining generation code.

You can download the latest model weights and view setup instructions on the HY-World-2.0 Hugging Face page.