Nvidia’s controversial DLSS 5 arrives September 3rd and requires serious GPU horsepower

Nvidia has revealed DLSS 5, the latest version of its controversial AI image enhancing technology, at Gamescom. After a backlash earlier this year, Nvidia is finally showing the tech in person. The message from Nvidia is clear: it's all about developer control. During a demonstration, a character in the Nvidia Zorah demo was shown sitting in a room, with DLSS 5 altering the image to a significant degree. The results were beautiful, yet ugly, impressive, and troubling all at once.

The character looked far more photorealistic than in previous reveals, but the image was nowhere near as brash as expected. Two sliders, "structure intensity" and "tone intensity," were set around the midway mark, resulting in a good-looking character. Developers have more tools to play with, including a masking feature to selectively enhance specific items and the ability to switch AI models on and off. The fine-tuning abilities of DLSS 5 are designed to be developer-friendly.

However, pushing DLSS 5 too far can result in an unwanted, gut-level reaction. Model B, in particular, made some people wince when the sliders were pushed up. The HDR-like shadows became overblown, the skin textures too shiny, and the facial features too keenly drawn. While models A and C looked good at various settings, model B made the characters look substantially different from the original, similar to AI beauty filters.

Nvidia's point is that developers, the artists behind the game, choose how far to push the technology. If it oversteps the mark, they can adjust the settings and rein it back in. Some Nvidia staff members questioned whether DLSS 5 can alter facial geometry, but arguments turned into a philosophical debate about the functional similarity of the result. In the end, the longer the demos continued, the clearer it became that DLSS 5 can create stunning imagery when handled delicately, and overblown results when pushed too far.