A pre-release version of Nvidia's Deep Learning Super Sampling (DLSS) technology, specifically DLSS 5, has reportedly leaked and is being experimented with by users across a range of games. The results, while not always optimized for the titles they are applied to, have produced visually striking and often peculiar outcomes.

Unlike previous iterations of DLSS that primarily focus on upscaling lower-resolution images to higher resolutions while maintaining fidelity to the original art, DLSS 5 appears to incorporate AI image generation. This feature aims to enhance the artwork it processes, with some comparing its capabilities to the results of using prompts like "photorealism" with AI image models.

One widely circulated image, humorously dubbed "Claude Strife," allegedly showcases DLSS 5 applied to Final Fantasy 7 Rebirth with extreme settings. This version of Cloud Strife exhibits an intensified realism, with pronounced skin textures and altered facial expressions, deviating significantly from the game's established art style.

Beyond the comedic, DLSS 5 implementations have also been observed to impart an aging effect on characters, accentuating details in skin, jawlines, and expressions. For instance, a Starfield character with otherwise flat textures was reportedly rendered with more human-like depth and detail when DLSS 5 was applied. A Starfield character with and without Nvidia's DLSS 5. Many examples of these experiments can be found in online communities dedicated to discussing PC gaming technology.

Opinions on these altered visuals are divided. While some users express delight at the potential for enhanced realism and character detail, others are reportedly aghast at the perceived "AI look" and the departure from original artistic intent. This divergence in reception mirrors broader discussions surrounding generative AI in creative fields.

Even when DLSS 5 succeeds in adding volume and texture to otherwise flat elements, such as faces, the results can carry a distinct artificial quality. This "AI look" has been described as resembling studio-lit photographs of models stretched over game characters, leading some to feel the output appears "fake" despite its technical complexity. The technology's early, unoptimized leaked state suggests that future official implementations may offer more refined and potentially less jarring results, but the underlying AI generation process appears to leave an inescapable stylistic signature.