Nvidia has provided more details about DLSS 5 after its initial announcement drew criticism from gamers and industry observers.
Critics raised concerns that the technology applied heavy beauty-filter effects to character faces and allowed artificial intelligence to interfere with developers’ original art direction. Reports that the early version required a second graphics card also caused concern.
During a follow-up keynote at SIGGRAPH, Nvidia focused on the level of control developers and artists will have over DLSS 5 rather than its ability to produce photorealistic graphics.
One of the main complaints about the original DLSS 5 demonstration was that the graphics appeared overly smooth and resembled filters commonly used on social media platforms.
Nvidia said artists will have full control over how the technology affects a game’s visuals.
Developers will not have to apply AI processing to an entire scene. They will be able to exclude individual characters, faces, objects and background elements or apply separate settings to them.
Structural Intensity and Tone Intensity controls will let developers adjust how strongly DLSS 5 alters the appearance of various visual elements.
Nvidia will also provide three separate AI models that developers can assign based on their requirements.
Another concern was that DLSS 5 could generate incorrect details on low-polygon geometry or replace carefully designed visual elements.
Nvidia said DLSS 5 does not generate game scenes from the beginning. Instead, it uses internal data provided by the game engine to preserve the artist’s original design.
Traditional rasterisation and ray tracing will continue to build the complete frame first. These systems will define the geometry, lighting reference points, motion vectors and surface normals.
DLSS 5 will then enhance more complex lighting elements.
Nvidia described the technology as a third category of AI rendering, alongside image upscaling and reconstruction systems and frame-generation technology.
Generative video models often process several frames together, which can introduce input delay. Such latency would create problems in fast-paced games such as Counter-Strike 2.
Nvidia said DLSS 5 will process the current frame using the game engine’s existing motion vectors. It will not examine future frames before producing the image.
By avoiding multi-frame batch processing, Nvidia aims to prevent additional generative latency and reduce temporal shimmering.
The original DLSS 5 demonstration reportedly required two high-end graphics cards due to its heavy computational demands.
Nvidia now says the final technology will run on a single RTX graphics card.
The company has reduced the original generative model into a smaller, single-step diffusion model. Nvidia designed the updated version to use video memory more efficiently and operate without a multi-GPU setup.
This change would make DLSS 5 more accessible to players who do not own multiple high-end graphics cards.
The SIGGRAPH presentation addressed the main concerns surrounding DLSS 5, including unwanted visual changes, possible interference with developer artwork, latency, and demanding hardware requirements.
However, the final performance and visual quality will only become clear when the planned update arrives in the fall.
The technology could offer richer game graphics without producing the artificial appearance that concerned users after the initial demonstration.
Get the latest tech news, telecom insights, and product launches wherever you prefer.
Add ProPakistani to Preferred Sources and see more of our stories in Google Search and Top Stories.