Another real time engine has joined the growing list of platforms capable of rendering gaussian splats.
The open source Wicked Engine, developed by graphics programmer János Turánszki, recently introduced native Gaussian splat import and rendering as part of its latest update. The addition allows users to load splat models directly into the engine’s editor, expanding the engine’s modern rendering pipeline to support 3DGS alongside traditional mesh assets.
Wicked Engine is a C++-based real-time rendering engine designed for performance focused graphics applications, games, and research experimentation. It runs across Windows, Linux, macOS, and modern consoles while supporting modern GPU features such as DirectX 12, Vulkan, ray tracing, HDR rendering, and variable rate shading.
With this update, gaussian splats can now be imported directly from PLY files and rendered inside the engine. The renderer effectively treats splats as a specialized particle system. Instead of camera facing billboards, each splat carries its own orientation and appearance parameters. The latest update also replaces the engine’s previous GPU bitonic sort with a radix sort implementation. Because splats must be sorted back to front each frame for correct transparency blending, sorting can dominate rendering cost. In testing with the famous bike scene, the new radix sort reduced sorting time from roughly 38 milliseconds to about 6.8 milliseconds.
At present, splat assets are limited to static scenes, and dynamic lighting support has not yet been added. Sorting across multiple independent splat models also remains a future improvement, as the current implementation assumes a unified particle system during rendering.
Integration of Gaussian splatting into full game engine environments has continued to accelerate over the past year. The addition of splats into Wicked Engine opens the door for hybrid scenes where traditional meshes and radiance field reconstructions coexist inside the same engine. Wicked Engine is released under the MIT License and is available on GitHub.
