Zorah is a large-scale Nvidia RTX neural rendering tech demo powered by Unreal Engine 5, designed to exhibit the newest developments in real-time rendering. Initially offered at CES 2025 and additional refined for GDC, the demo is now publicly obtainable as a 108 GB obtain.
It highlights key improvements in geometry processing, lighting, character rendering, and neural shader applied sciences, and is meant for builders and researchers working with next-generation graphics pipelines.
Observe: The obtain is packaged as an Unreal Engine challenge, not an executable. Requires you to put in Unreal Engine first.
The challenge runs on the Nvidia RTX department of Unreal Engine 5 (NvRTX) and is optimized for the GeForce RTX 5090, using options from the Blackwell structure. Zorah consists of scenes composed of practically 500 million triangles, over 30,000 supplies, greater than 2,000 particle lights, and 1,500 high-resolution textures.
These property are rendered in actual time utilizing neural rendering applied sciences, together with RTX Neural Shaders, DLSS 4 with body technology, and a transformer-based AI mannequin for dynamic content material technology.
Neural rendering strategies utilized in Zorah embrace real-time technology of lighting, textures, hair, and supplies immediately throughout the shader pipeline. Supporting applied sciences corresponding to ReSTIR PT and DI (for path-traced world illumination and direct illumination), RTX Mega Geometry, Neural Materials networks, and Neural Texture Compression additional improve visible constancy and rendering effectivity.
The demo operates completely with real-time path tracing – no rasterization – offering an correct preview of how future video games and functions would possibly leverage neural strategies to enhance each high quality and efficiency.
Zorah is designed for high-end growth techniques and requires a contemporary RTX GPU, ideally an RTX 5090, to attain real-time efficiency.