Nvidia has revealed DLSS 5, but they don’t seem to have gotten the reception they wanted.
As they explain in their blog post, the latest iteration of DLSS uses an AI model. As opposed to AI image generators, this model doesn’t take data from other artists and creative works.
Instead, its inputs are the color and motion vectors that appear in each game frame. The model is then trained to add photorealistic lighting. Presumably, the model is also trained with the data Nvidia has collected since they launched DLSS over a decade ago.
This reveal received backlash almost immediately, with images like our feature image making the rounds. Nvidia has already made a statement declaring that it is not an AI filter.
From what we’ve seen, DLSS 5 doesn’t really yassify every image or frame. Rather, the AI model is inconsistent, which makes it the more surprising that Nvidia saw it fit to reveal it in this way.
Game developers have full control over DLSS 5 effects such as intensity, color grading, and just masking it off. Suffice to say that Nvidia and devs will have to convince the public that this AI implementation is worthwhile.