DLSS 4.5 integration in Unreal Engine 5.7 introduces advanced super resolution with improved image quality and expanded multi-frame generation for smoother performance, alongside flexible model selection and Nvidia Streamline support for enhanced frame pacing and debugging. The discussion also covers technical setup best practices, limitations like VR support and editor viewport constraints, and highlights the plugin’s adaptability and source code availability for developer customization.
The discussion centers around the DLSS 4.5 integration in Unreal Engine 5.7, featuring Gabby Sullins, a principal software engineer specializing in DLSS within the Unreal Engine context. DLSS 4.5 introduces enhancements primarily in super resolution and multi-frame generation. The super resolution improvements include new second-generation transformer models that significantly boost image quality, reduce artifacts, and enhance effects like anti-aliasing and motion clarity. Meanwhile, the multi-frame generation feature has been expanded from 5x to 6x, with dynamic scaling to optimize performance by computing only the necessary frames, improving smoothness even when the game’s CPU is a bottleneck.
The conversation also delves into the technical aspects of installing the DLSS plugin in Unreal Engine. The recommended approach is to place the plugin in the engine’s marketplace folder to avoid obscure packaging errors, especially for blueprint-only projects that lack C++ source code. Source-engine users have more flexibility and can place the plugin in other engine plugin directories since they build the engine themselves. This distinction is important because Unreal Engine 5 has stricter requirements for plugin placement compared to Unreal Engine 4, and following these guidelines ensures smoother integration and fewer technical issues.
Regarding model selection within the project settings, DLSS offers various models tailored to different quality levels and content types. While the default models are optimized for most use cases and future updates, developers have the option to override these settings to better suit their specific content or to address artifacts. This flexibility allows pro users and developers to fine-tune image quality and performance. Additionally, features like ray reconstruction have multiple presets, and switching to newer models can yield noticeable improvements in global illumination quality, especially in path-traced scenes.
The discussion highlights the role of Nvidia Streamline, which acts as a service layer enabling features like Reflex and Frame Generation to work cohesively. Streamline manages frame pacing and communication between these technologies. Developers can configure Streamline settings in the project settings panel, including enabling debug overlays that provide detailed runtime information useful for diagnosing issues with Frame Generation or Reflex. These overlays complement blueprint-based debugging tools and offer different levels of insight, helping developers understand and optimize DLSS and Streamline behavior within their projects.
Finally, the conversation touches on some limitations and historical challenges of DLSS integration in Unreal Engine. For example, Frame Generation cannot be displayed directly in editor viewports due to DirectX swap chain constraints and requires a standalone window. VR support for DLSS is currently limited by engine constraints and would require changes from Epic Games. The most memorable technical challenge was adapting to Unreal Engine 5.1’s major changes to the upscaler interface, which shifted control of screen percentage from plugins to the engine itself to better support multiple upscalers. Overall, the DLSS plugin is designed to be flexible and extensible, with source code available for developers who want to customize or extend its functionality further.