YVO3D moves to WORLD, its version 3, which generates a complete 3D environment from a single image.

YVO3D WORLD generates modular, multi-object 3D environments from a single image for Unreal Engine 5, reducing weeks of production work to minutes.

YVO3D, a 3D model generation platform for game developers, artists, and studios, is evolving its offering from the creation of isolated objects to that of entire environments. Until now, the tool transformed text or an image into photorealistic models up to 8K, with automatic retopology, clean UV unwrapping, and usable materials, exportable to Unreal Engine, Unity, Blender, and Maya.

The new feature, named WORLD, builds a structured and modular environment from a single image. Rather than a single, fixed mesh, the scene is broken down into distinct objects that can be moved, replaced, or reused, making it directly manipulable in a production pipeline, especially within Unreal Engine 5. Two modes coexist: a fully automatic generation and a hybrid approach offering finer control. The user can also inject their own library of elements for the system to assemble.

The produced scenes achieve great complexity, sometimes several hundred million polygons, with clean topology, continuous PBR materials—meaning calculated for physically realistic rendering—and reconstructed depth even in areas absent from the original image. The stated goal is to reduce work that previously took several weeks to just a few minutes. The generation of unitary models is already available in beta, and WORLD is expected to arrive in early access for Premium and Enterprise accounts. The platform is already being used in production by Tencent.