Nvidia's DLSS 5 neural rendering: what it is, how it works, and why gamers are skeptical

Nvidia is trying to rehabilitate the image of its DLSS 5 neural rendering technology after March's backlash, when gamers saw screenshots of Resident Evil Requiem character Grace Ashcroft rendered with the system and immediately declared it "AI slop." Now, with a September 3rd launch on the horizon and leaks giving modders early access to the tech, the company is messaging that the system respects artistic intent and hands developers full control over how—or whether—it modifies their game graphics.

What DLSS 5 actually does#

DLSS 5 functions as a generative AI filter applied to 2D rendered game frames after the 3D engine finishes its work, similar to a camera filter on your smartphone. Rather than fundamentally altering game geometry, it manipulates light, shadow, color, and surface detail to make existing shapes appear more refined or photorealistic.

The technology can enhance contact shadows that ground characters to surfaces, add realistic light reflections through hair and fabric, and sharpen details on nearby objects—but it does not create new shapes, generate text, or hallucinate objects that weren't there. Nvidia claims the results are deterministic, meaning the same scene should render identically each time, unlike generative AI images that vary with each pass.

However, the performance cost is substantial. Testing shows DLSS 5 can cut framerates roughly in half, and it currently requires Nvidia's latest and expensive RTX 50-series graphics cards.

Developers—not gamers—hold the controls#

A critical selling point is granular customization. Developers can adjust sliders that control how aggressively the technology affects "structure" (shadows and definition) and "tone" (light and color), with three distinct AI models available at launch labeled "default," "natural," and "cinematic."

Crucially, they can also isolate which parts of a scene receive the enhancement—character faces and skin, foliage, objects like silverware or food—using automated masks or custom selections. This means studios could theoretically leave character faces untouched while enhancing environmental lighting and detail.

The flip side: Nvidia did not confirm whether gamers will receive any controls to dial back the effect themselves, only stating that developers have full authority over the final implementation.

Why leaked mods look so bad#

The horrifying results circulating online—where characters appear aged decades or unnatural—typically stem from intensity settings cranked to 80 percent or higher. The leaked mod shows 's Lara Croft weathered by deepened wrinkles and creases at maximum settings. At 40 percent intensity, the same character appears subtly enhanced rather than transformed.