Earlier this week, Nvidia debuted DLSS 5, a brand new generative AI upscaling know-how that purports to permit builders to get even nearer to realizing their creative visions by putting weird AI porn slop faces on in style characters like Grace from Resident Evil Requiem. Individuals obtained mad. Actual mad. Now Nvidia is responding by doubling down on slop-scaling.
CEO Jensen Huang was requested concerning the on-line backlash at this week’s GPU Know-how Convention by Tom’s Hardware. “Properly, to begin with, they’re utterly improper,” he responded, arguing that full management of the tech stays with the sport builders.
“DLSS 5 fuses controllability of the of geometry and textures and every thing concerning the recreation with generative AI,” Huang stated. “It’s not post-processing, it’s not post-processing on the body stage, it’s generative management on the geometry stage.”
He claimed builders who’re on board with the tech, together with at Capcom and Bethesda, can “fine-tune the generative AI” to form their recreation’s visuals how they need. DLSS 5 provides a brand new generative AI-fueled layer of constancy however doesn’t take away “creative management.”
Reporter: Some are involved NVIDIA’s DLSS 5 will make video games look worse or homogeneous.
In our Q&A immediately, Jensen Huang responded:
“They’re utterly improper… you may advantageous tune the generative AI to your creative model… if you need cartoon, toon shader, manufactured from glass… it’s up… pic.twitter.com/cTjzUnZoiW
— Justin Ryan ᯅ (@justinryanio) March 17, 2026
Huang could also be responding to the way in which most individuals have interpreted and digested the tech on-line, which is basically as a type of “slop filter” that provides an uncanny layer of hyper-fidelity to video games primarily based on generic coaching fashions relatively than the imaginative and prescient of the unique artists.
That’s due no less than partly to how demo movies present earlier than and after comparisons that make characters’ faces in video games like Starfield and Hogwarts Legacy appear to be yassified stunt doubles. Huang’s argument is that that is being calibrated on the developer stage relatively than through a post-processing algorithm.
I’m undecided that logic will make a lot of a distinction to any of the followers at present horrified by the tech’s early outcomes. In any case, what’s worse? DLSS 5 as an inexpensive slop-face filter or DLSS 5 as a fusion of generative AI into the basic geometry that defines how a recreation appears and feels?
DLSS 5 may do wonders for extra generic particulars like environmental lighting, however the instruments are by their very nature considerably random and unpredictable. Most individuals’s solely query up to now about DLSS 5 is whether or not Nvidia can assure an choice to maintain it turned off.
