NVIDIA Releases RTX Remix Runtime Code

NVIDIA has open-sourced the runtime components of the RTX Remix modding platform, which allows existing classic PC games based on the DirectX 8 and 9 APIs to add support for rendering with simulated light behavior based on path tracing, improve the quality of textures using machine learning methods, connect user-prepared game resources (assets) and apply DLSS technology to realistically scale images to increase resolution without losing quality. The code is written in C++ and is open under the MIT license.

TX Remix Runtime provides pluggable DLLs that allow you to intercept game scene processing, replace game resources during playback, and integrate support for RTX technologies such as path tracing, DLSS 3, and Reflex into your game. In addition to the RTX Remix Runtime, the RTX Remix Platform also includes the RTX Remix Creator Toolkit (still announced), which is powered by the NVIDIA Omniverse and allows you to create visually enhanced mods for some classic games, attach new assets and lights to recycled game scenes, and use methods machine learning for processing the appearance of game resources.

NVIDIA Releases RTX Remix Runtime Code

Components included in RTX Remix Runtime:

  • Capture and replace modules responsible for intercepting game scenes in USD (Universal Scene Description) format and replacing original game resources with modernized ones on the fly. To capture the stream of rendering commands, a substitution of d3d9.dll is used.
  • Bridge, which translates 32-bit render engines to 64-bit ones to remove limitations related to available memory. Prior to processing, Direct3D 9 calls are converted to the Vulkan API using the DXVK layer.
  • A scene manager that uses the information coming in through the D3D9 API to create a representation of the original scene, track game objects between frames, and set up the scene to apply path tracing.
  • Path tracing engine that renders, processes materials and applies advanced optimizations (DLSS, NRD, RTXDI).



Source: opennet.ru

Add a comment