This article was originally published on link to post
Nvidia has announced its next addition to the growing suite of DLSS options, which already includes impressive upscaling, ray reconstruction for ray-traced workloads, and frame generation. DLSS 5 is being teased as the biggest leap forward yet, with Nvidia introducing an AI-powered model that changes lighting and material properties at a per-pixel level. However, the response to the announcement has been overwhelmingly negative so far.
Equated to the introduction of a programmable graphical shader in 2001, Nvidia is leveraging a real-time neural rendering model that takes in color and motion vectors from a single in-game frame and generates a wholly new one with photorealistic lighting and material changes. Nvidia notes that these changes are anchored to a game’s assets, indicating that the model doesn’t inherently change the geometry of a character’s facial model, for example, but instead just reshapes the lighting that around it, providing a new level of detail.
The technology was predominantly showcased using character models from games such as Resident Evil Requiem, Starfield, and The Elder Scrolls IV: Oblivion Remastered. The results are striking, but for polarizing reasons. Lighting is indeed photorealistic, but seems to also drastically alter the look of some characters to

0 Comments