NewsTradingSentimentEventsCommunityBriefing
Tech

Nvidia's Leaked DLSS 5 Reshapes Old Games with Unintended Results

By Tech Desk · 2026-09-20 · 3 min read
A rain-slicked city street at night with wet asphalt reflecting neon lights
Illustration: Tradingbird

A leaked update to Nvidia's rendering technology is causing a stir by altering the visual style of classic titles, producing results that range from photorealistic improvements to uncanny distortions.

For nearly three weeks, the gaming community has been divided over a leaked version of Nvidia's DLSS 5. The technology, which uses artificial intelligence to enhance graphics, is being described by some as a violation of artistic intent. Critics argue it imposes a generic, machine-generated look on games, stripping away the unique visual styles developers crafted. However, the feature has also garnered massive attention on video platforms, where millions of views suggest a strong public interest in seeing how AI can modernize older titles.

To understand the impact, I tested this leaked software on three well-known games from different eras. The goal was to see if the technology could genuinely improve immersion or if it would break the visual identity of the games. The results were inconsistent, showing that the AI does not treat all artistic styles with equal neutrality. In some cases, it enhanced the atmosphere significantly; in others, it created a disturbing disconnect between the character design and the rendered output.

Character Faces Become Uncanny

Testing on The Witcher 3 revealed a significant flaw in how the system handles non-human characters. The protagonist, Geralt, has a distinctive, mutated appearance that is central to his identity. Under the new rendering, his face appeared waxy and overly defined, falling into what is known as the uncanny valley. Instead of looking more realistic, he looked like a stranger wearing his face, which detracted from the immersion. This suggests the AI struggles to interpret stylized or non-standard human features, potentially misidentifying them as errors to be corrected.

In contrast, the same game’s human characters, such as Triss, benefited from the update. Her face appeared more flattering and natural, indicating that the system likely relies on training data dominated by conventional human aesthetics. This inconsistency confirms that the technology is not neutral regarding art direction. It favors specific types of visual representations over others, which raises valid concerns about whether developers can still maintain their unique artistic vision when using such tools.

Atmospheric Environments Gain Realism

The technology performed much better when applied to environments, particularly in Batman: Arkham Knight. This title features a dark, rainy cityscape that aligns well with the strengths of modern rendering. The AI enhanced the wet asphalt, rain effects, and lighting, making the city of Gotham feel heavier and more tangible. The result was a more fleshed-out Batman who fit the ominous atmosphere of the setting. For environments that rely on realistic physics and lighting, the update provided a clear visual upgrade without conflicting with the game's established tone.

This duality presents a complex trade-off for developers and players. While the AI can significantly boost the visual fidelity of backgrounds and lighting, it risks fundamentally altering the character designs that define a game's identity. The system appears to prioritize a generic standard of realism, which may not always align with the creative goals of the original developers. This creates a scenario where the same tool can be a major improvement for one part of a game and a detrimental change for another.

Artistic Intent Faces AI Bias

The core issue highlighted by these tests is the lack of transparency regarding how these models are trained. Nvidia has provided limited information on the data used to build this version of DLSS. The bias toward conventional human features suggests that the AI is learning from a narrow subset of visual data. This means that games with stylized, animated, or highly customized character designs may suffer the most. The technology is not simply adding detail; it is reinterpreting the image through a specific, perhaps unintended, aesthetic lens.

As reported by GN auto tech/entertainment, the community remains split on whether this represents the future of gaming or a step backward for artistic control. The leaked version of DLSS 5 demonstrates that artificial intelligence in graphics is not a neutral utility. It carries its own set of preferences and limitations. For players, the choice to use such features will require a careful assessment of whether the visual gains outweigh the potential loss of the original artistic intent.

Based on reporting by xda-developers.com, compiled by the Tradingbird desk.

Read next

More in Tech

More from the Tech desk

All desk stories