Nvidia’s newest SIGGRAPH 2026 keynote confirmed off extra in-depth management within the fingers of builders for the much-criticized DLSS 5, together with completely different ‘ranges’ of controllable era from inside the instrument for “a brand new stage of realism.”
DLSS 5, which was revealed earlier in 2026 and was additional clarified to successfully function a 2D AI filter was blasted on-line, with Nightfall developer New Blood referring to the tech as “generative AI bullsh*t.”
Quick-forward to July 2026, and Nvidia’s SIGGRAPH 2026 keynote on Tuesday, July 21 revealed extra in regards to the controversial characteristic. Throughout the presentation, Edward Liu, Director of Utilized Deep Studying Analysis, states that with the instrument, a recreation continues “constructing the world precisely as the sport has authored it. And the era turns into the discovered stage afterwards to counterpoint its look.”
Discussing how a generated picture can output completely different outcomes as a result of there’s “plenty of room for the mannequin to improvise,” he provides that it would not work for one thing like real-time rendering, as what’s offered to the viewer within the following body is not in a position to be picked. “So we’d like some situation that’s extra constrained and rigorous.” This course of introduced three notable challenges: Preserving creative intent, frame-by-frame streaming, and real-time 4k.
From what’s mentioned subsequent, precisely how the era will seem in-game can be managed by the builders themselves, together with choices to “fine-tune depth on a Per-Body Foundation.” This contains components like subsurface scattering, reflections, and lighting. Total, DLSS 5 is alleged to not be “AI changing graphics, [but] AI extending graphics. The identical manner programmable shaders and ray tracing have prolonged graphics earlier than.”
Whether or not or not you are workforce DLSS 5 or not, you will not have for much longer to see it in motion, because it’s delivery later this 12 months in Fall 2026.






