At GTC, Nvidia announced DLSS 5, a whole new version of the company’s Deep Learning Super Sampler upscaler tech, coming later this fall (Q3 2026) for Nvidia RTX 50 series graphics cards.
Nvidia is touting this as “the most significant breakthrough in computer graphics since the debut of real-time ray tracing in 2018,” referring to the early days of the Nvidia RTX cards.
DLSS 5 can do more than just upscale a video game frame from a low resolution to a higher one, or generate in-between frames. DLSS 5 can now add photoreal lighting and materials on the pixels themselves, creating “a new level of photoreal computer graphics previously only achieved in Hollywood visual effects.”
Nvidia founder and CEO Jensen Huang says “DLSS 5 is the GPT moment for graphics—blending handcrafted rendering with generative AI to deliver a dramatic leap in visual realism while preserving the control artists need for creative expression.”
And that’s the crux of the controversy. To gamers, this is really is a “GPT moment” (derogatory) in that to their eyes, DLSS 5 is adding an AI slop filter, “yassifying” characters by adding uncanny levels of photorealism.
One of the DLSS 5 demo examples shows Resident Evil Requiem’s protagonist Grace looking much more “realistic” with DLSS 5 on, mostly because her face seems to have changed not just by additional lighting, but also added details that was not there before like the addition of makeup.
The examples using Hogwarts Legacy shows the main character, a teenager, having more detailed wrinkles and other pleasant-looking facial details that ended up making him look way older.
The AI model that is adding these details isn’t hallucinating them, at least that’s what is reported by those who had seen the real-time demos up close.
“It’s not a face filter,” Ryan Shrout reported in his impressions of DLSS 5. The tech also improves lighting of the environments, which may be the actual good use case for the tech rather than sloppifying the faces.
Game developers also have “detailed controls for intensity, color grading and masking” which they can control when integrating DLSS 5.
The issue is that gamers hate AI slop. And any effect that resembles that, which DLSS 5’s examples showcase with the face comparisons, will be an instant turn-off. In fact, it has, based on the reactions now out on the internet.
And there are good reasons why gamers are so anti-generative-AI: costs of PC parts including SSD storage and sticks of RAM has now ballooned up due to shortage as supply is being prioritised to AI data centers. Graphics cards offering hasn’t been exciting, or as affordable (Nvidia RTX 50 series cards MSRP ranges from RM1,499 to a whopping RM10,329) as companies like Nvidia focus priority in the AI boom (some would say AI bubble). And the amount of issues generated by generative AI, from the abundance of fake/questionable content to the environmental impact of creating slop motion graphics have all made gamers all the more weary of generative AI.
DLSS, in its previous forms, is one of the more acceptable use cases of generative AI. Yes, it requires the deep learning model to be fed an abundance of video game footage to train it. But once that’s done, the upscaling work is all local, crunched by dedicated cores on a GPU of a PC, and the result of improved performance and/or visuals are more or less worthwhile.
DLSS 5, however, is crossing a line now that it also can now add unexpected new details instead of just upscaling and generating frames. The tech may be impressive, but it’s deployed at the wrong time and circumstances, hence the controversy.
These games will support DLSS 5 when it launches:
- AION 2
- Assassin’s Creed Shadows
- Black State
- CINDER CITY
- Delta Force
- Hogwarts Legacy
- Justice
- NARAKA: BLADEPOINT
- NTE: Neverness to Everness
- Phantom Blade Zero
- Resident Evil Requiem
- Sea of Remnants
- Starfield
- The Elder Scrolls IV: Oblivion Remastered
- Where Winds Meet