Jensen Huang Says DLSS 5 Critics Are "Completely Wrong" Over Artistic Control Concerns

More

18 March 2026 11:41

TL;DR

Nvidia CEO Jensen Huang pushed back directly against DLSS 5 criticism at the company's annual conference, stating critics are "completely wrong" and that the technology gives developers direct generative control at the geometry level rather than applying AI as post-processing over finished frames.
DLSS 5 drew backlash after its reveal, particularly over side-by-side comparisons showing an AI sheen on character models in Resident Evil: Requiem that critics argued undermined the game's intended visual style and artistic direction.

Jensen Huang is not a person who hedges when he thinks he's right. So when Tom's Hardware asked him about the DLSS 5 backlash at Nvidia's annual conference, the response was characteristically direct.

"Well, first of all, [the critics] are completely wrong." That's the opening line. Not "we understand the concern" or "let me provide some context." Completely wrong. It sets a tone, and then Huang actually made the case.

What He's Actually Arguing

The criticism of DLSS 5 centred on a familiar fear: that generative AI applied to game visuals overrides the artistic choices developers made deliberately. The Resident Evil: Requiem comparison became the focal point, with the DLSS 5 version of Grace appearing to have an AI-smoothed quality that detracted from the original's intentional aesthetic.

Huang's counter-argument rests on a technical distinction that matters if accurate. "DLSS 5 fuses controllability of the geometry and textures and everything about the game with generative AI," he said. The key word is "fuses." He's arguing this isn't AI operating on top of a finished frame, it's operating within the render pipeline itself, at the geometry level.

"It's not post-processing, it's not post-processing at the frame level, it's generative control at the geometry level," Huang said. "All of that is in the control – direct control – of the game developer. This is very different [from] generative AI; it's content-control generative AI. That's why we call it neural rendering."

If that distinction holds in practice, it's significant. Traditional upscaling, including earlier DLSS versions, takes a lower-resolution rendered frame and reconstructs detail around it. The artistic decisions are already baked in before upscaling touches anything. Huang is claiming DLSS 5 works differently, with developers able to fine-tune the generative AI as a tool rather than accepting what it produces wholesale.

Why the Scepticism Persists

The problem is that the demo didn't show that level of control clearly. Side-by-side comparisons where the AI version looks noticeably different from the intended output are exactly what you'd expect from post-processing, not from a developer-directed geometry-level process. Saying the developer has control and demonstrating it convincingly are different things.

More:NLC Left Without Tournament Organiser as League ApS Steps Down Amid Payment Crisis

Tags: Nvidia

Share:

Sources:

Report mistake Submit news

Related news

View All

Cloud9 Partners with Nvidia

Cloud9 has announced a new collaboration with NVIDIA. The partnership will see the integration of NVIDIA GeForce RTX 40 Series...

Sponsorships

Sep 06, 2024

Nvidia is Growing

Nvidia has achieved the largest single-day increase in market value ever recorded on Wall Street. On Wednesday, the California-based chip...

Business

Aug 01, 2024

Nvidia's Q4 Financial Results Revealed

Nvidia announced its fourth fiscal quarter results for fiscal 2024, ending on January 31, 2024, with revenue reaching $22.1 billion....

Business

Feb 22, 2024

US Government Warns NVIDIA

The United States Government is restricting NVIDIA from selling chips to China. The company has been redesigning the chips with...

Drama

Dec 04, 2023

Nvidia Office Raided in France

French authorities raided Nvidia offices in France in antitrust investigation. The French competition authority announced this week that it is...

Drama

Sep 29, 2023

Jensen Huang Says DLSS 5 Critics Are "Completely Wrong" Over Artistic Control Concerns

More

18 March 2026 11:41

Tags: Nvidia

TL;DR

Nvidia CEO Jensen Huang pushed back directly against DLSS 5 criticism at the company's annual conference, stating critics are "completely wrong" and that the technology gives developers direct generative control at the geometry level rather than applying AI as post-processing over finished frames.
DLSS 5 drew backlash after its reveal, particularly over side-by-side comparisons showing an AI sheen on character models in Resident Evil: Requiem that critics argued undermined the game's intended visual style and artistic direction.

Jensen Huang is not a person who hedges when he thinks he's right. So when Tom's Hardware asked him about the DLSS 5 backlash at Nvidia's annual conference, the response was characteristically direct.

"Well, first of all, [the critics] are completely wrong." That's the opening line. Not "we understand the concern" or "let me provide some context." Completely wrong. It sets a tone, and then Huang actually made the case.

What He's Actually Arguing

The criticism of DLSS 5 centred on a familiar fear: that generative AI applied to game visuals overrides the artistic choices developers made deliberately. The Resident Evil: Requiem comparison became the focal point, with the DLSS 5 version of Grace appearing to have an AI-smoothed quality that detracted from the original's intentional aesthetic.

Huang's counter-argument rests on a technical distinction that matters if accurate. "DLSS 5 fuses controllability of the geometry and textures and everything about the game with generative AI," he said. The key word is "fuses." He's arguing this isn't AI operating on top of a finished frame, it's operating within the render pipeline itself, at the geometry level.

"It's not post-processing, it's not post-processing at the frame level, it's generative control at the geometry level," Huang said. "All of that is in the control – direct control – of the game developer. This is very different [from] generative AI; it's content-control generative AI. That's why we call it neural rendering."

If that distinction holds in practice, it's significant. Traditional upscaling, including earlier DLSS versions, takes a lower-resolution rendered frame and reconstructs detail around it. The artistic decisions are already baked in before upscaling touches anything. Huang is claiming DLSS 5 works differently, with developers able to fine-tune the generative AI as a tool rather than accepting what it produces wholesale.

Why the Scepticism Persists

The problem is that the demo didn't show that level of control clearly. Side-by-side comparisons where the AI version looks noticeably different from the intended output are exactly what you'd expect from post-processing, not from a developer-directed geometry-level process. Saying the developer has control and demonstrating it convincingly are different things.

More:NLC Left Without Tournament Organiser as League ApS Steps Down Amid Payment Crisis

Share:

Sources:

Report mistake Submit news