Nvidia just gave developers the closest thing to a control panel for DLSS 5 that the tech has had so far. At SIGGRAPH 2026, the company explained that devs will get two main sliders to control the intensity of DLSS 5’s AI effects: “structural intensity” and “tone intensity.” The big question since DLSS 5 was announced in March is whether it was going to feel like a one-size-fits-all “filter” slapped over a game scene. Nvidia’s answer is more specific than a promise, and it is also a bit of a letdown: two sliders is what it’s putting on the dev side.

Here is what those two sliders actually do, and why Nvidia thinks they matter. Nvidia says the “structural intensity” slider modulates the high-frequency detail of the frame. In Nvidia’s examples, that includes ambient occlusion, subsurface scattering, and reflections. The “tone intensity” slider modulates the low-frequency detail of the frame, covering “expressive” elements such as lighting and overall tone. Nvidia also emphasizes an important boundary condition for artistic control: it is “not changing the geometry in any way, and it is uplifting the image.” So the controls are not about rewriting a scene, they are about steering how the AI enhances different parts of the image.

To understand why this matters beyond aesthetics, rewind to how DLSS 5 is built. Nvidia previously confirmed that the tech has to work with a traditional 2D frame and motion vectors for each AI frame. That confirmation essentially validated the “filter” interpretation from earlier demos. In that framing, intensity controls are the most direct lever studios have to dial how strong the enhancement is, but they are not a full artistic steering wheel. That is why the headline implication can feel like a disappointment to some teams: many expected more granular controls than a simple on-off mask.

Still, Nvidia’s current strategy is not random. The company’s described pipeline is designed to keep things “quick and snappy” for rendering by making the model “does not need to peek ahead,” instead using a render and motion vectors to create what is essentially an AI filter. When your input constraints are tightly defined (2D frames plus motion vectors), the practical output controls tend to map onto what the model can reliably adjust. Nvidia is telling devs that the model preserves artistic intent thanks to “internal buffers,” which is a key reason it may only expose these two intensity sliders instead of more knobs.

The most important nuance for production planning is that Nvidia is not stopping at global controls. Nvidia says it is “building more” controls for developers, and it is “working with our partners to incorporate their feedback.” On top of the two-slider framework, game devs will be able to mask different objects in a game scene, changing the structural and tone intensities for each independently. That is a big deal for practical implementation, because it turns two sliders into a more composable toolkit. For example, a studio could push background objects toward stronger structural enhancement but keep characters at a lighter touch. This is exactly the kind of “art-directable realism” teams tend to want when balancing quality targets, player perception, and performance.

Nvidia also provided a concrete illustration through a demo by a creative artist introduced as “Gaff.” The point of the example was that increasing structural intensity can add realism without forcing a major shift in overall tone. Nvidia notes that subsurface scattering can make a significant difference, and that Gaff says this “would be incredibly expensive to do in traditional rendering.” The operational takeaway is straightforward: where classic rendering approaches are costly, AI-enhancement intensity tied to specific image frequency bands could be a way for studios to get closer to premium visuals without paying the full traditional compute bill.

As for rollout timing and how studios should think about the roadmap, Nvidia positions this update carefully. Nvidia says this new addition will “extend the existing pipeline” and “does not replace them.” And it notes that, as of now, the company is still working on other controls. The implication for teams waiting to fully integrate DLSS 5 is that the current control surface is likely stable enough to build around, but it is not the final word. Nvidia also says DLSS 5 will launch this fall, which matters for production timelines: studios will need to lock up integration details, QA plans, and tuning passes before the broader feature set is necessarily in place.

So the strategic stake is not just “two sliders.” It is how small, explicit controls can change a studio’s adoption risk. If the AI behaves consistently within “structural” and “tone” constraints, teams can standardize presets, reduce tuning cycles, and more easily compare visual deltas across hardware. If they need deeper control later, the promise that Nvidia is “building more controls” is the bridge. For peers making similar decisions, the message is clear: DLSS 5 is being framed as an extensible pipeline enhancement with an intentionally bounded first set of knobs, plus object masking to give art direction a foothold.