Nvidia CEO Jensen Huang has pushed back hard against the growing criticism of DLSS 5, the company’s new AI-powered neural rendering technology unveiled at GTC 2026 this week. When Tom’s Hardware asked him about complaints that the technology was making games look worse or stripping out artistic intent, Huang didn’t mince words: “Well, first of all, they’re completely wrong. “
His defence centres on how DLSS 5 actually works under the hood. Huang explained that the technology fuses a game’s existing geometry, textures, and motion vectors with generative AI—it doesn’t override the artist’s work, it builds on top of it. Crucially, he stressed it isn’t a post-processing filter slapped on a finished frame. “It’s generative control at the geometry level,” he said, adding that developers can still fine-tune the model to match their intended visual style.
Why gamers think DLSS 5 looks like an ‘AI filter’
The backlash kicked off almost immediately after Nvidia released demo footage showing DLSS 5 applied to games including Resident Evil Requiem and Hogwarts Legacy. Critics—including several game developers—said the results looked airbrushed, overly shiny, and devoid of character. Thomas Was Alone developer Mike Bithell said the technology seemed designed for when you “absolutely, positively, don’t want any art direction. ” Gunfire Games concept artist Jeff Talbot was blunter, calling it “just a garbage AI filter. “
It’s worth noting that the demo Nvidia showed has its sliders maxed out and runs across two RTX 5090s—one dedicated entirely to the DLSS 5 model. The version that ships this fall will be optimised for a single GPU, and will presumably be more nuanced.
Huang points to Ray Tracing as a precedent for doubters
Whether DLSS 5 follows the same arc remains to be seen. Nvidia has time—the feature isn’t launching until fall 2026—and more controlled demos will likely follow before then.
