DLSS 5 and its 3D-guided neural rendering are on track to be the next big thing in real-time graphics and game visuals. Like the arrival of the original DLSS and ray tracing with the GeForce RTX 20 Series, what we're seeing here is the first version of something that will mature and evolve over years and generations of GeForce RTX hardware. Still, the impressive photorealism it brings to NBA 2K27, the first game to support DLSS 5, is already up there in terms of "wow" factor.
And it all comes down to two things: DLSS 5's model, designed to deliver photorealism and cinematic lighting in a way that would be impossible even with modern path tracing on a GeForce RTX 5080. And second, how it has been trained to 'preserve artistic intent.' On top of that, NVIDIA is giving developers detailed tools and controls over how it's applied and how it looks in any given scene.
Although its controversial debut several months ago led many to believe DLSS 5 worked like other generative AI tools that create images and videos, that's not the case. NVIDIA confirms that a generative model simply isn't suited for real-time graphics, which is why DLSS 5 was built and trained to take information from a rendered frame and enhance it.
DLSS 5 can identify characters in a game, from faces to silhouettes to clothing and costumes. DLSS 5 can also identify objects in a scene, materials like metal, wood, and fabric, and determine light sources and direction. In NVIDIA's real-time Zorah tech demo, we see DLSS 5 recognize a table, a barrel, and roses, alongside human elements like hands and a face.
With a DLSS 5 development team that includes computer graphics veterans who have worked on Pixar films and visual effects for years, alongside AI experts, DLSS 5 is also trained to identify camera placement and composition, and even the mood. This real-time information not only informs DLSS 5's output but also the controls developers have for tuning. This includes the ability to mask each object and apply different levels of intensity for the model and tone-mapping/color, and the ability to choose from three differently weighted models.
With these controls, the overall effect can be subtle but still additive and transformative or more prominent in turning a rendered frame into one that looks photoreal. In NBA 2K27, developer Visual Concepts uses DLSS to turn the game's photorealistic ambition into reality through a combination of subtle changes that add up.
Frequently Asked Questions
Open a question for an answer from TweakTown's coverage of this news, or ask your own below.
How does DLSS 5 identify and treat different in-game objects like faces, clothing, and materials?
What developer controls does DLSS 5 provide for masking objects and adjusting intensity or tone mapping?
Can DLSS 5 be tuned to produce subtle enhancements rather than overt photoreal effects?
Does DLSS 5 function like generative AI image models, or how is it different for real-time graphics?
Have a question about this content not listed here? Ask below and TweakBot will answer it.
This includes more definition and detail in things like hair, facial hair, and eyebrows. Skin looks more real, with visible pores and beads of sweat. The improved shadows and occlusion add immensely to the detail, too, with clothing looking more realistic alongside little things like shadows forming around a player's eyes.
