Nvidia’s latest SIGGRAPH 2026 keynote showed off more in-depth control in the hands of developers for the much-criticized DLSS 5, including different ‘levels’ of controllable generation from within the tool for “a new level of realism.”
DLSS 5, which was revealed earlier in 2026 and was further clarified to effectively serve as a 2D AI filter was blasted online, with Dusk developer New Blood referring to the tech as “generative AI bullsh*t.”
Fast-forward to July 2026, and Nvidia’s SIGGRAPH 2026 keynote on Tuesday, July 21 revealed more about the controversial feature. During the presentation, Edward Liu, Director of Applied Deep Learning Research, states that with the tool, a game continues “building the world exactly as the game has authored it. And the generation becomes the learned stage afterwards to enrich its appearance.”
Discussing how a generated image can output different results because there’s “a lot of room for the model to improvise,” he adds that it doesn’t work for something like real-time rendering, as what’s presented to the viewer in the following frame isn’t able to be picked. “So we need some condition that is more constrained and rigorous.” This process brought three notable challenges: Preserving artistic intent, frame-by-frame streaming, and real-time 4k.
From what’s discussed next, exactly how the generation will appear in-game will be controlled by the developers themselves, including options to “fine-tune intensity on a Per-Frame Basis.” This includes elements like subsurface scattering, reflections, and lighting. Overall, DLSS 5 is said not to be “AI replacing graphics, [but] AI extending graphics. The same way programmable shaders and ray tracing have extended graphics before.”
Whether or not you’re team DLSS 5 or not, you won’t have much longer to see it in action, as it’s shipping later this year in Fall 2026.
