Today at Gamescom 2025, NVIDIA unveiled updates to NVIDIA RTX neural rendering and NVIDIA ACE generative AI technologies that enable developers to deliver cutting-edge, scalable gaming experiences and reach broader audiences. The updates are explained in this post and include:
- Expanded integration options for DLSS 4 and NVIDIA RTX Kit technologies
- New models and inference backends to build AI-driven experiences across any GPU vendor
- Tool improvements for optimizing graphics and AI workloads
- Powerful cloud-based solutions like GeForce NOW inside Discord
This technology is behind upcoming titles including Resident Evil Requiem, Borderlands 4, and The Oversight Bureau.
More DLSS 4 integration options for developers
DLSS 4 is a revolutionary suite of neural rendering technologies that uses AI to boost FPS, reduce latency, and improve image quality. Originally announced with the launch of the GeForce RTX 50 Series at this year’s CES, DLSS 4 supported titles has grown to over 175. With NVIDIA Blackwell coming to GeForce NOW cloud gaming, members also benefit from DLSS 4 with Multi Frame Generation, and other AI-powered features—allowing developers to deliver their games in maximum graphical fidelity to players, even if they don’t have the latest hardware.
For developers using custom engines, the Streamline SDK is available with the latest updates and fixes. For Unreal Engine developers, the DLSS 4 plugin is now available in Unreal Engine 5.2 to 5.6.
Expanded support for NVIDIA RTX Kit
NVIDIA RTX Kit is a suite of technologies that allow developers to train and deploy AI directly within shaders, build fully path traced games with real-time performance, and make photorealistic characters.
The Microsoft DirectX 12 Agility SDK preview is now available with support for Cooperative Vectors. This functionality provides developers direct access to RTX Tensor Cores from within DirectX shaders, resulting in substantial performance improvements and memory savings in real-time graphics applications. It paves the way for more realistic and efficient AI rendering techniques.
NVIDIA has updated the RTX Neural Texture Compression SDK with a D3D12 path for Cooperative Vectors, which provides evaluation, texture compression, and engine integration tools alongside a library for decompression. This technology dramatically reduces memory usage of high-quality textures without sacrificing quality.
New improvements to other RTX Kit technologies include:
- RTX Neural Shaders SDK: Introduces a new sample for converting from SlangPy (a library designed to make calling GPU code written in Slang easy) to C++. Slang is used to train and deploy tiny neural networks for shaders.
- RTX Character Rendering SDK: Adds animation support to the digital human sample, making it easier to review RTX Hair and Skin implementations.
- RTX Texture Filtering SDK: Brings new optimizations that improve the quality of filtered textures when magnified.
- RTX Global Illumination SDK: Received bug fixes and user experience updates.
To get started, download RTX Kit technologies and read the tutorial guide for tips.
RTX Kit technologies are made easier for Unreal Engine developers through the NVIDIA RTX Branch of Unreal Engine (NvRTX) 5.6. The latest state-of-the-art path tracing algorithm, ReSTIR PT, enables an infinite number of lights, accurate indirect illumination, and significantly improved mirror reflections. ReSTIR PT is currently only available in the NvRTX branch of Unreal Engine 5.
RTX Mega Geometry accelerates bounding volume hierarchy (BVH) builds for cluster-based geometry systems and makes it possible to ray trace full quality Nanite geometry. RTX Mega Geometry is now available as an experimental feature in NvRTX 5.6, giving developers an opportunity to experiment with the technology prior to its production release.
The latest NvRTX branch updates include:
- RTX Mega Geometry (experimental)
- Support for cluster operations that enable real-time ray tracing of full quality Nanite geometry
- Lighting
- More stable ReGIR sampling using a stationary grid. Decoupling the grid from camera movement achieves a more stable Next Event Estimation (NEE) sampling camera under motion.
- GI approximation for strand-based hair, powered by a dedicated path tracing core optimized for hair strand pixels.
- Per-light diffuse and specular intensity scaling, allowing more flexible per-light tuning.
- A simplified implementation of Lumen-style reflections, reducing integration complexity.
- Performance and debugging
- Significant reduction in ReSTIR PT/GI shader permutation count, easing build complexity and improving maintainability.
- Enhanced debugging features, including improved RTXDI Offline Path Tracer and a new Pixel Print utility to inspect register-level behavior.
- Updated DLSS plugins that improve
-d3ddebugand-rhivalidationcompatibility.
Finally, a new scene has been added to the NVIDIA Zorah technology sample (Figure 3). It showcases the latest ReSTIR DI algorithm, which is now available in the RTX Dynamic Illumination SDK and NvRTX 5.6. The multibounce lighting from the dynamically shadowed area brings new realism and richness to the environment. To explore and learn more, download and compile the scene.
To learn more and get started, download NVIDIA RTX Branch for Unreal Engine 5.6 and watch the technical overview webinar.
New ACE models and inference backends
NVIDIA ACE is a suite of technologies that brings game characters to life with AI. It enables a new generation of living breathing worlds, filled with knowledgeable companions, dynamic agents, and conversational actors.
Today, the NVIDIA Riva Automatic Speech Recognition (ASR) model is available for on-device inference. This model provides high word accuracy with real-time performance (90-100 ms time to first text) with word-boosting and streaming functionality for English transcription.
This model was showcased in The Oversight Bureau by Iconic Interactive. You play as Candidate 404, trapped in an underground reconditioning facility. Every day you’re forced to participate in simulations as the Bureau attempts to reprogram your mind and turn you into a more compliant member of society. The player interacts with the game using speech, which the ACE model then transcribes into text.
Developers can get started with Riva ASR on-device through the NVIDIA In-Game Inferencing (NVIGI) SDK. GPU-optimized for high performance, the SDK is a plugin-based inference manager designed to simplify the integration of AI models such as Riva ASR into C++ gaming and interactive applications. The latest update, NVIGI 1.2, provides Direct3D support to inference language models across all GPU vendors as DLLs.
Improve graphics performance with NVIDIA Nsight developer tools
NVIDIA Nsight Graphics is a standalone tool for designing, debugging, and optimizing games and other graphics applications. In version 2025.4 the Graphics Capture activity has been promoted from beta to production-ready.
Capture acquisition is streamlined to capture one or more frames immediately to persistent disk storage, enabling more flexible workflows for both interactive and batch usage of frame capture. API support in Graphics Capture includes NVIDIA RTX Kit features such as RTX Mega Geometry and RTX Hair, as well as D3D DirectStorage and Vulkan Device-Generated Commands.
Download Nsight Graphics to get started.
NVIDIA and Discord partnership opens new opportunities for game developers
NVIDIA and Discord have teamed up to announce a new integrated experience that brings seamless, cloud-powered game discovery to millions of players—right inside Discord—powered by GeForce NOW cloud gaming.
The integrated experience enables players to jump straight into a game they see their friends playing on Discord with no downloads, installs—all without having the game or GeForce NOW installed on their PC. A limited-time trial of the GeForce NOW Performance experience powers the gameplay, streaming at up to 60fps in 1440p, showcasing the quality and power of NVIDIA cloud technology. The first integration is being demoed behind closed doors at Gamescom with the first game to take advantage of the feature, Epic’s popular free-to-play Fortnite.
For game developers and publishers, instant access where gamers already are means faster engagement and wider reach. GeForce NOW inside Discord also bypasses hardware limitations and connects game developers with players in the communities they already play, watch, and talk in.
Connect your game to where players already are—frictionless, social, and powered by GeForce NOW. Developers who are interested can reach out to Discord at partnerships@discordapp.com.
Get started with new RTX neural rendering tools
You can deliver cutting-edge, scalable gaming experiences and reach wider audiences with the latest updates to NVIDIA RTX neural rendering and NVIDIA ACE—from expanded integration options for DLSS 4 and RTX Kit to new inference backends for building AI-driven experiences.
To learn more, watch the webinar replay about the latest NVIDIA RTX Branch for Unreal Engine 5.6, and check out NVIDIA resources for game developers.
