Although in recent years Nvidia has become the company with the largest market capitalization thanks to its in-demand AI chips, the American company has been known for decades for your graphics cards for computers. This Monday, Jensen HuangCEO of Nvidia, presented the next version of the technology DLSS (Deep Learning Super Sampling). Launched in 2018 with GPU series RTX 2000DLSS upscale video games from a lower resolution using machine learning and even, as the technology has matured, interpolating frames created by AI to have a higher performance of images per second. This allows many users to play at resolutions that exceed the capabilities of their GPU if they try it natively and do it with great visual fidelity. Until it has arrived DLSS 5.
Huang announced the new version, which will be released in the fall, and presented several examples of the technology during the GTC 2026 event, with games such as Resident Evil Requiem either starfield. Since then, criticism has not stopped, both from users and developers who accuse Nvidia of having created a ‘AI filter’ which, although it adds more details, modifies the artistic direction of video games that use it. This has not pleased Huang, who has described the criticism as ‘completely wrong’.
What is DLSS 5
Nvidia describes DLSS 5 as ‘a real-time neural rendering model’ capable of ‘delivering a new level of photorealistic computer graphics that until now was only achieved in Hollywood visual effects’. Nvidia CEO explained at GTC 2026 that the technology combines ‘generative AI’ with ‘artisanal rendering’ to give ‘a drastic leap in visual realism while maintaining the control artists need for creative expression’.
Unlike current generative video models, which Nvidia notes are ‘difficult to control precisely and often lack predictability,’ DLSS 5 uses a game’s internal color and motion vectors ‘to imbue the scene with photorealistic lighting and materials anchored to the source 3D content and consistent from frame to frame’. According to the company, that internal game data helps the system ‘understand complex scene semantics, such as characters, hair, fabrics, and translucent skin, along with ambient lighting conditions such as front light, backlight, or cloudy sky.’
All of this sounds great on paper and it is true that DLSS 5 offers a higher level of detail than the original, as can be seen in the videos, but at what cost? Too much, for many developers and players.
Developer Reviews
Mike Bithellcreator of Thomas Was Alonehas stated that the technology seems designed ‘for when you absolutely and adamantly don’t want any artistic direction in your gaming experience’. AND Jeff TalbotSenior Conceptual Artist at Gunfire Gamesbelieves that ‘in each shot the artistic direction was eliminated to add meaningless “details.” Every DLSS 5 shot looked worse and had less personality than the original. This is nothing more than a garbage AI filter‘.
Dave Oshryfounder and CEO of New Blood Interactivepoints out that DLSS 5’s ‘AI shit’ ‘is really depressing’ and lamented that future generations ‘won’t even know that this looks “bad” or “wrong,” because for them it will be normal.’
Bethesdaone of the many large publishers that Nvidia mentioned as an early partner of DLSS 5, has justified on social networks that these videos are ‘a very preliminary look, and our art teams will continue to adjust the lighting and the final effect to make them look how we think works best in each game. All of this will be under the control of our artists and will be totally optional for players.‘.
DLSS 5 OFF // DLSS 5 ON
(image or embed)
— Tuba Zef (@tubazef.com) March 16, 2026, 8:59 p.m.
Huang’s response
Nvidia has noted in the comments of the DLSS 5 trailer on YouTube that ‘it’s not a filter’ and? ‘Developers have full, detailed artistic control over DLSS 5 effects to ensure they maintain the unique aesthetic of their game’. That includes the ability to adjust the intensity and color grading or to completely disable masking in ‘places where the effect should not be applied’.
Huang also responded to the criticism during a Q&A session with the press at GTC 2026. ‘Well, first of all, they are completely wrong. The reason is that, as I have explained very carefully, DLSS 5 merges the ability to control geometry, textures and everything related to the game with generative AI,’ he said in response to Tom’s Hardware.
The CEO insisted that the generative capacity of technology ‘it doesn’t change artistic control’ and that the developer can fine-tune how the generative AI acts to suit the style of the game: ‘All of that is under the direct control of the game developer. This is very different from generative AI; is generative AI with control over content. That’s why we call it neural rendering.’
For now, Nvidia has confirmed that Bethesda, Capcom, Hotta Studio, NetEase, NCSoft, S-GAME, Tencent, Ubisoft and Warner Bros. Games are collaborating with the company on the use of DLSS 5.
Nvidia has also confirmed that the DLSS 5 demos shown needed two RTX 5090 to runone to render the game and another dedicated exclusively to the DLSS 5 model. However, he assured that the final version will only require a single GPU.
The company has not yet announced real-world performance figures or hardware compatible with this technology, but the need for 2 RTX 5090, Nvidia’s most powerful gaming GPU, suggests that the launch It will be aimed at premium graphics cards, as happened with the first DLSS 8 years ago.