3 mins
SIGGRAPH 2026: Nvidia DLSS 5 explained — what changes on your screen
Wondering what all the buzz is about at SIGGRAPH? Here’s Nvidia DLSS 5 explained in plain language: NVIDIA just showed new developer tools that shape how DLSS 5 alters a frame — letting teams mask specific scene elements and tune structural and tonal intensity so the final image better matches a game’s art direction.
That means less guesswork from the AI and more hands-on steering from artists. For players, it translates to visuals that trend toward photoreal lighting and materials while staying loyal to a game’s intended style.
Quick Summary
- DLSS 5 adds per-object masking and two intensity controls so studios can keep the game’s look on-model.
- There are three real-time DLSS 5 models (A/B/C) that developers can mix per scene or character.
- Launch remains targeted for Fall 2026; DLSS 5 runs in real time up to 4K on a single GPU.
What you’ll actually see on screen
DLSS 5’s model analyzes each frame’s color and motion vectors, then adds a photoreal lighting pass that respects the underlying 3D content. Expect more convincing skin (subsurface scattering), fabric sheen, hair highlights, and light bounces that feel grounded — with greater temporal consistency than simple post effects.
Crucially, developers can protect a game’s visual identity. If the original art calls for flatter, stylized materials or specific color grading, they can dial DLSS 5 back on those elements or exclude them outright, preserving the intended mood.
Image credit: NVIDIA
A Quick Explanation
Structural Intensity governs material/geometry cues — things like surface definition, reflections and microdetail. Tone Intensity influences color and contrast behavior, aligning the final grade with the scene’s lighting so characters and environments don’t drift from the art style.
NVIDIA’s new dev controls: masking + two dials
The updated toolkit lets creators mask any number of characters or objects to apply more (or less) DLSS 5 influence per element. That’s helpful for hero faces, branded props, or materials that must remain stylized. Once masked, Structural and Tone Intensity sliders let artists refine how strongly the lighting/material pass shows up.
Think of it as targeted enhancement instead of a full-frame filter. Because the controls are per-object aware, a studio can keep HUD props or a hand-painted asset intact while still boosting realism elsewhere in the shot.
For PC Gamers
Don’t expect these creator sliders in your video menu. You’ll likely see a simple DLSS 5 toggle and quality preset. The artistry happens in development builds, so differences between games will reflect how each studio chose to mask and tune the effect.
The three real-time DLSS 5 models, explained
NVIDIA presented three real-time models — informally labeled Model A, Model B, and Model C. Each leans toward a different enhancement intensity and computational cost, and all are designed for single-GPU operation. Studios can mix models within a scene: a subtler choice for fast action, a stronger pass for cinematic close-ups.
In practice, that gives teams a flexible palette. A character’s face might use one model with conservative Structural Intensity to retain sculpted detail, while the environment uses another model with a slightly richer Tone setting to lift indirect light and material response.
Image credit: NVIDIA
Worth Noting
DLSS 5’s pass is anchored to the game’s content, not a free-form prompt. The model works from the frame’s color and motion data, aiming for deterministic, temporally stable results that still leave room for a studio’s art and lighting choices.
How this compares to FSR and XeSS right now
AMD’s FSR and Intel’s XeSS focus on upscaling and frame generation — reconstructing resolution and inserting intermediate frames to raise performance. They don’t add a separate photoreal lighting/material pass to the image. DLSS 5 goes beyond upscaling by enhancing lighting and materials while keeping output tied to the game’s 3D scene.
That difference could matter in side-by-side shots. If a developer leans into DLSS 5’s controls, you might notice cleaner global illumination, more natural skin and fabric response, or fewer “plastic” highlights — all while staying true to the game’s look. The flip side: quality will vary by title, since the tools enable discretion rather than a one-size preset.
What’s still unknown
We’ve seen the controls and NVIDIA’s tech demos, but the real test will be shipping games on varied content and engines. How often studios ship per-object masks for every tricky case — hairlines, decals, very stylized materials — will shape the day-one experience. Expect image-quality debates to continue, especially in motion.
For PC Users
DLSS 5 is slated for Fall 2026 on GeForce RTX. Early titles NVIDIA has named include Starfield, Resident Evil Requiem and Assassin’s Creed Shadows, with more to follow. Update drivers, and check in-game DLSS 5 presets when patches land — studios decide how aggressive the effect is per scene.
Bottom line: DLSS 5 is shifting from a bold idea to a controllable toolset. If developers embrace the new masking and intensity tuning, players should get sharper lighting and materials without losing the art that makes each game distinctive.
Image credit: NVIDIA
