The irruption of DLSS-5 It marks one of those moments when the video game industry feels like it's entering a new era. It's not just about increasing the frame rate or stretching the resolution, but about letting the generative AI in video games intervene directly in how each scene looks of a real-time game.
Presented during the conference Nvidia GTC 2026This new version of the well-known AI supersampling system aims to be the biggest graphical leap since the arrival of ray tracingThe company maintains that, thanks to a model of neural renderingVideo games will be able to approach the visual finish of cinema without forcing GPUs to render every detail in a classic way.
What is DLSS 5 and why does Nvidia call it the “GPT moment of graphics”?
To date, Nvidia's DLSS has been primarily associated with gain performance: render at a lower resolution, upscale with AI, and, in the latest versions, generate additional frames to increase smoothness. With DLSS 5, the approach changes fundamentally. The goal is no longer just to produce more pixels, but transform the quality of those pixels with the help of generative AI.
As Nvidia has explained, DLSS 5 takes as its input structured game engine data —color, motion vectors, geometry, materials, and depth— and uses them as if they were a “prompt” for a neural model, according to analysis on How video game developers use AIThat model interprets the entire scene and It injects advanced lighting and photorealistic materials over the image that the game has already rendered.
Jensen Huang, the company's CEO, himself defined the technology as “the GPT moment of the charts”The comparison is not accidental: just as large language models combine structured and generative information to produce text, DLSS 5 blends traditional rendering with AI to increase the level of visual detail without completely breaking the control that developers have.
On paper, the proposal aims Closing the gap between real-time graphics and Hollywood special effectsWhile a film can afford minutes or hours of calculation per frame, a PC game has to render each image in about 16 milliseconds if it wants to maintain 60 FPS. That's where AI comes in as a shortcut.
How DLSS 5's real-time generative AI works
Instead of simply scaling an image or "inventing" intermediate frames, DLSS 5 intervenes in every frame to enrich itThe simplified workflow would look something like this:
- The game engine generates a base version of the scene with its original textures, materials and lighting.
- DLSS 5 receives the motion vectors, color information, geometry, and depth of that frame.
- The AI model “understands” what is on the screen —skin, hair, fabrics, water, metal, smoke, etc.— and It applies different treatments to each type of material..
- The result is a reconstructed image with More complex lighting and more believable materials, generated in real time and consistently between frames.
The key is that the model doesn't work blindly: it's anchored to the actual game dataThis allows the changes to appear not as a simple random filter, but as a layer of detail linked to the scene. Nvidia insists that the system is deterministic and stable between frames, something fundamental to avoid flickering or annoying artifacts in a moving game.
In practice, DLSS 5 is capable of recreating effects that until now were reserved for film: subsurface dispersion in the skin (the light penetrating slightly and then exiting again), subtle glimmers in fabrics and textilesmore realistic reflections in water and metalOr a interaction of light with hair much richer. All this, the company points out, without modifying the original geometry or textures of the game.
According to Nvidia, the system works up to a resolution of 4K in real timeIn the first public demonstrations, the company has even used two RTX 5090 cardsOne for running the game and another dedicated exclusively to the AI model. However, the plan is for the commercial version to run on a single next-generation RTX GPUHowever, that makes it clear that the hardware bar won't be exactly low.
Photorealistic lighting and materials: the great visual leap
The change that DLSS 5 brings is most noticeable in the lighting and the treatment of materialsUntil now, ray tracing and path tracing techniques had already raised the bar, but at a huge performance cost. Nvidia's proposal is to delegate some of that work to a generative model that, supported by the scene's physical data, Complete the lighting without having to calculate each ray of light in a classic way.
The model is trained to recognize in a single frame Elements such as characters, clothing, translucent surfaces, or lighting conditions (backlighting, front lighting, cloudy skies, interiors with mixed lighting, etc.) are considered. From there, it adjusts how the light behaves in each area: shines on fabrics, subtle reflections, softer or harder shadows depending on the source, and an overall sense of volume that, in the demos, is quite reminiscent of the finish of a pre-rendered scene.
Nvidia emphasizes that The original assets and textures are not touched.Although in some comparisons it's hard to believe that only the lighting has been changed. In titles like Resident Evil: RequiemEg The faces appear to change beyond just the lighting, something that part of the community has already pointed out as a possible point of friction.
In any case, the company insists that the improvement is, by design, a layer of lighting and materials applied over the original contentThe model fills in the gaps that real-time rendering cannot cover without spiking resource consumption, with the promise that the result will still be recognizable to developers and players.
This approach also opens the door to something Nvidia has been hinting at for some time: in many cases, the vast majority of the screen's pixels will no longer be "original"but rather generated or reinterpreted by AI from a relatively modest base. The company has even mentioned that, with technologies of this type, more than twenty out of every twenty-four pixels can come from the model's work.
Creative control and doubts about artistic direction
One of the questions that most concerns studios and players is to what extent DLSS 5 respect the artistic direction of each game. After all, we're talking about a neural network deciding in real time how a scene, already designed by a human team, should look; this concern was reflected in debates such as the video game actors' strike.
Nvidia says that creators will continue having the last wordDLSS 5 is integrated through the framework NVIDIA Streamline and offers developers specific controls for intensity, color gradation, and masks to limit where and how AI is applied. In theory, that should allow the effect to be tailored to each project, from a photorealistic title to one with a more stylized look.
Even so, part of the community—including analysts from specialized media outlets and players active on forums and social networks—has already begun to talk about a certain effect “post-processing filter”Some compare the result to running the game through an AI beautification layer that, in specific scenes, It could eat away at part of the original intention in terms of color, contrast or atmosphere.
The expression has gained traction on platforms like Reddit and X (formerly Twitter). “AI slop” to refer to images that, according to its critics, appear over-processed, as if the scene had been excessively retouched. The controversy revolves around a simple issue: Does it improve the experience or blur the style of the game?
For now, the debate is more theoretical than practical: until DLSS 5 is released to the public and players can enable and disable it on their own systems, it will be difficult to measure accurately. how much real margin of control do the studies have? and what impact it will have on the day-to-day development process.
Compatible games and support from major publishers
Beyond the technical promises, Nvidia has presented a list of titles and studios that will accompany the launch of DLSS 5. Among the first confirmed games are: productions with great appeal on PC and console, many of them with a strong presence in Europe.
The company has showcased the technology in action in titles such as Starfield, Hogwarts Legacy, Resident Evil: Requiem, Assassin's Creed Shadows and remaster of The Elder Scrolls IV: OblivionIn the demos, the comparisons between DLSS 4 and DLSS 5 focused precisely on shadows, reflections on wet surfaces and volumetric light effects, which gain naturalness and geometric coherence.
Regarding studios and publishers, the initial list includes names such as Bethesda, Capcom, Ubisoft, Warner Bros. Games, Tencent, NetEase, NCSOFT, Hotta Studio or S-GAMEThat is, both large Western players —with a strong presence in Spain and the rest of Europe— and Asian giants with increasing influence in our market.
In addition to the titles already mentioned, Nvidia has announced compatibility with other projects such as AION 2, Delta Force, Phantom Blade Zero, NARAKA: BLADEPOINT, Where Winds Meet y others in developmentThe idea is that DLSS 5 won't be limited to a few "showcase" games, but will gradually reach a wider audience. extensive catalog throughout 2026 and 2027.
For European gamers, this early adoption by publishers with a strong presence in the region means that most major PC releases They could offer DLSS 5 as a graphics option as soon as compatible hardware is widespread enough.
Hardware requirements and availability: who will be able to use DLSS 5
Officially, DLSS 5 is planned for fall 2026 and it will arrive, first of all, at the next-generation GeForce RTX graphics cardswith the RTX 50 series in its sights. Nvidia speaks of a multi-year development process and continuous optimization until launch, but for now remains silent on the specific requirements.
The data that has been made public paints a demanding picture: some demonstrations have been run with two RTX 5090 GPUsOne focused on the game and another dedicated exclusively to the AI model. Although the company claims that in its laboratories DLSS 5 now works on a single cardEverything suggests that only the most powerful GPUs will be able to take advantage of it reliably at high resolutions.
In the console market, the situation is uneven. PlayStation 5 and Xbox Series X|S use graphics chips from AMD and they lack the specific AI cores that Nvidia uses in its RTX cards, so They will not be able to use DLSS 5 natively. Instead, they will continue to rely on their own solutions such as FSR (AMD) or internal technologies such as PSSR in the case of the future PS5 Pro.
The exception is the next one Nintendo Switch 2which does integrate Nvidia hardwareThis opens the door for Nintendo's console to potentially benefit, with some nuances, from part of the DLSS ecosystemIt remains to be seen to what extent a portable device—with a much smaller energy budget—will be able to implement advanced DLSS 5 features, or if it will be limited to lighter versions.
In Europe, where the park of Gaming PCs with RTX GPUs While DLSS 5 adoption is particularly widespread in countries like Germany, France, the UK, and Spain, its success will largely depend on the rate at which gamers upgrade to RTX 50 series or higher. For many users with current setups, the decision will be whether it's worth investing in a high-end GPU just to access this visual leap.
Between technical promise and controversy over “excessive AI”
As is often the case with Nvidia announcements, the initial impact of DLSS 5 has been a mix of technical admiration and skepticismOn the one hand, the demos shown at GTC 2026 effectively display scenes with a level of detail and lighting that a few years ago would have seemed difficult to imagine in real time.
For another, Not everyone in the community has received the announcement with enthusiasm.Some players feel that certain screenshots and videos look too "polished," to the point that the characters' faces and the overall atmosphere... They deviate from what the studios showed in their original trailers.That is the origin of the aforementioned term “AI slop”, which is used in a derogatory way to describe AI-generated images that appear artificial or over-processed.
There are also reasonable doubts about the impact on developmentIf DLSS 5 is going to become a central piece of the graphics pipeline, studios will likely need to design their games thinking from the beginning about how AI will interveneThis could help optimize resources—by delegating certain levels of detail to the model—but it also creates dependence on proprietary technology and a specific provider.
Comparisons with the leap that the programmable shaders Ray tracing and other technologies are inevitable. Then, as now, there was initial reluctance, doubts about performance, and debates about whether the change was worthwhile. The difference is that, this time, we're not just talking about new GPU instructions, but about letting a statistical model... interpret each frame of the game.
Until DLSS 5 is widely available on PC and, to a lesser extent, on platforms like Switch 2, what does seem clear is that the conversation surrounding the Generative AI in video games It will continue. Between those who see it as the only way to keep moving forward with realism without skyrocketing costs, and those who fear it will end homogenizing the appearance of many titles or diluting the artisanal work; figures like Kojima have expressed caution regarding artificial intelligence.
With DLSS 5, Nvidia is openly betting on a future in which Most of the visuals in a video game are mediated by AIRelying on engine data but reconstructing the scene in real time, the technology promises to bring games closer to the visual level of cinema, provided the appropriate hardware is available. However, it also forces us to rethink what we mean by "real-time graphics" and to what extent we want to delegate how we see and play to a neural model.
