NVIDIA showcases updated DLSS 5 with developer tools that help 'preserve artistic intent'

NVIDIA opened its keynote at the annual SIGGRAPH 2026 event in Los Angeles this week with an updated and in-depth look at DLSS 5. And compared to the announcement and debut from GTC earlier this year, which led to widespread backlash for the generative AI tool designed to add photorealistic lighting and detail to a scene, it was night and day.

2

In addition to providing a more technical look at how it works, the big thing with this presentation was showcasing how much control game developers will have over DLSS 5, and how they will be able to direct the look, feel, and intensity of the technology to "preserve artistic intent."

From a high level, this updated version of DLSS 5 includes three real-time models that developers can use, each presenting photorealistic lighting with different intensity. These models, dubbed Model A, Model B, and Model C, can be used on different scenes and characters, with the option to mix and match. Like DLSS Super Resolution, some of the models are more intensive and complex than others, but all are designed to run on a single GPU.

Unlike the initial DLSS 5 debut, the demonstration of the effect and developer tools was carried out on NVIDIA's own RTX tech demos built to showcase the latest in RTX Blackwell ray-tracing and rendering technologies. A smart move, but also one that makes sense as it allows the technology, models, and developer customization tools to take center stage on this updated look at DLSS 5.

When it comes to DLSS being rendered as part of a scene, NVIDIA notes that the entire traditionally rendered frame is taken into account, with DLSS 5's photorealistic lighting pass focused on complex lighting like sub-surface scattering, global illumination, and more. NVIDIA notes that the biggest challenge with DLSS 5 is adding material and object photorealism while maintaining the artistic look and visual identity of the source material.

DLSS 5 development tools offer in-depth customization, with the ability to mask characters and objects in a scene, with no limit, and then fine-tune the effect on each. The two main sliders are Structural Intensity and Tone Intensity, with each dictating different types of effects like reflections, and so forth. For those criticisms aimed at how DLSS 5 dramatically altered the look of characters during its initial debut, these sliders will allow developers to highlight characters and fine-tune the effect so it looks additive rather than transformative.

Frequently Asked Questions

TweakBot answers common questions about this news using TweakTown's own coverage from this page and related content from our archive. Tap a question to reveal the answer, or type your own below.

Question #1

What are the differences between DLSS 5's Model A, Model B, and Model C and when should developers use each?

DLSS 5 offers three real-time models (Model A, Model B, and Model C) that present photorealistic lighting at different intensities and varying computational complexity. Developers can apply different models to different scenes or characters and mix and match them, choosing higher-intensity or more complex models where needed and lighter models elsewhere; all models are designed to run on a single GPU. Developer controls like Structural Intensity and Tone Intensity sliders and per-object masking let teams fine-tune each model's effect to preserve artistic intent.

Answered

Question #2

How do the Structural Intensity and Tone Intensity sliders affect DLSS 5's photorealistic lighting pass?

The Structural Intensity and Tone Intensity sliders let developers control how DLSS 5’s photorealistic lighting pass is applied to scene elements, with each slider dictating different types of effects such as reflections and other material or object details. Using those sliders (and masking), developers can fine-tune and highlight characters or objects so the DLSS 5 output appears additive to the original art rather than transformative.

Answered

Question #3

Can developers mask multiple characters and objects with DLSS 5, and are there limits to masking?

Yes. DLSS 5 development tools let developers mask characters and objects in a scene with no limit, and then fine-tune the effect on each. The tools include Structural Intensity and Tone Intensity sliders to control different aspects of the effect.

Answered

Question #4

How does DLSS 5 integrate with a traditionally rendered frame and which lighting effects does it target (e.g., global illumination, sub-surface scattering)?

DLSS 5 takes the entire traditionally rendered frame into account and applies a photorealistic lighting pass on top of it. That lighting pass targets complex lighting effects such as sub-surface scattering and global illumination, among other material and object photorealism enhancements.

Answered

Have a question not listed here? Ask below and TweakBot will answer it.

Your questionAsk

Answering your question×

TweakBot is analyzing your question…

Could not get an answer

TweakBot is generating an answer

Read More

Was TweakBot helpful?Thanks for the feedback!

Naturally, that's what DLSS 5 ultimately can do: offset some of the hardware-intensive work required to deliver a full path-traced scene with cinematic or photorealistic detail that is transformative. The good news is that NVIDIA has seemingly taken all of the criticism to heart, as what we see here looks, well, impressive - especially from a developer control perspective. Although this was a technical presentation, from what we've heard, DLSS 5 is still on track for a Fall 2026 release.