Early this morning, almost every gaming community was discussing one topic—DLSS 5.
DLSS 5 is a new AI technology launched by NVIDIA during the GTC 2026 keynote speech early today. Its birth was described by Jensen Huang as the "GPT moment in computer graphics."
You might be wondering: "Is it really as exaggerated as Jensen Huang claims?"
So, please take a look at the GIF comparing the effects of enabling and disabling DLSS 5 in Resident Evil Requiem:
The effect after enabling DLSS 5 is nothing short of earth-shattering.
You can see that the texture of human skin and lip color are incredibly realistic, pupils accurately reflect highlights, and the material and lighting performance on other object surfaces have also seen leapfrog improvements. The realistic visual effect presented by the entire scene far exceeds real-time rendered images, resembling pre-rendered high-spec CG...
But most people's first reaction is: "Why does this look so much like 'AI face-swapping'?"
After seeing the before-and-after comparison above, you probably noticed one thing: the appearance of game characters seems to have changed significantly with the activation of DLSS 5. Grace, who originally looked somewhat youthful, was directly transformed from a workplace novice into an elite white-collar worker with distinct facial features, changing the entire art style.
Of course, in the provided DLSS 5 technical demo video, Grace is far from the only character that viewers felt was drastically altered.
Leon, nearing fifty, suddenly transformed from a weary,沧桑 old man into a literary Oppa from a Korean melodrama.
As for the protagonist in Hogwarts Legacy, he no longer looks like a young fifth-grade transfer student. The oily shine typical of middle-aged men is enough to spawn a large number of US Marines.
After watching the demo video, everyone realized that DLSS 5 is comparable to East Asian dark arts. Any character that goes through its hands will be transformed beyond recognition, to the point where even their own mother wouldn't recognize them.
Consequently, netizens began creating meme images based on this point, teasing the plastic-surgery-like effects of DLSS.
First and foremost, naturally, are the game characters from the "same root."
Originally, in Resident Evil Requiem, the "girl" who terrified me so much that I wanted to pull out my gun and shoot upon seeing her, instantly became voluptuous and appetizing after being drastically altered by the "netizen version of DLSS 5." After seeing the "girl" with DLSS 5 ON, I no longer wanted to pull out my gun to shoot, but I did want to pull out my gun to shoot.
As for the monster below, which looked difficult to fight before being beautified by DLSS 5, it still looks difficult to fight after beautification, but ultimately it's not impossible to defeat. However, my skills are too poor, so I'll leave it to the experts among you to tackle first.
Another character, Kratos, who treats gods like little girls and is quite formidable, can only be considered a little girl under the divine power of DLSS 5. After being tortured by DLSS 5, he lost his past stern and cold demeanor, replaced by a pair of passionate, fiery red lips.
After seeing these tough characters, let's change perspective and give it a "Final Fantasy" thought—what would happen if DLSS 5 were applied to characters who are already breathtakingly beautiful?
With the support of DLSS 5, Alice might truly transform into a younger version of Auntie Cai.
If we changed Grace's rendering style, she could even become the famous Western internet goblin xQc.
As for Ellie in The Last of Us, why did she become so "smooth" in the TV series?
Naturally, this is because TV series production doesn't use DLSS 5, so the detail presentation and lighting shaping dropped a notch.
That's right, our DLSS 5 is just that awesome—even if you're a tree, as long as you follow big brother DLSS 5, you can end up looking presentable and respectable.
Besides messing with relatively realistic game visuals, some netizens ventured into more abstract territories.
What kind of chemical reaction would occur if the realistic effects of DLSS 5 were applied to abstract game visuals? For example... Minecraft?
Thus, you can see the image below: Steve, let me ask you, why is your head so square?
Additionally, the answer to one of the biggest unsolved mysteries in Let the Bullets Fly has finally come to light—why doesn't Zhang Mazi look like Ma Bangde in the photo?
It turns out Zhang Mazi forgot to turn on DLSS 5 when taking the photo.
Finally, people even started making jokes at Jensen Huang's expense.
You don't say, I think this mysterious leather-clad man still retains his charm.
After seeing so many "internet open-source DLSS 5" meme images, it's not hard to find that beyond simple teasing, some memes also contain an emotion—opposition and resistance towards DLSS 5 technology.
Whether it's drastically altering characters beyond recognition, transforming monsters to have intense sexual appeal (?), or replacing 2D characters with 3D models, these memes reflect players' concerns about DLSS 5 technology.
Besides memes, many people in the bullet comment pool and comment section of the technical demo video also expressed their dissatisfaction with DLSS 5, and indeed all AI technologies in the gaming field.
Some even fell into slippery slope speculations regarding the application of AI technology.
Among these emotions, there is natural public resistance to AI technology itself, but more importantly, there is concern that DLSS 5 technology might disrupt the original style and experience of games. The phenomenon of changing the artistic style of scenes and "face-swapping" characters in the demo video, along with the "AI taste" on the "new faces," are significant factors contributing to this concern.
Many people feel that DLSS 5 is merely an abuse of existing AI functions.
However, DLSS 5 doesn't work by crudely applying a uniform AI filter to all games. It integrates into the NVIDIA Streamline technology framework, allowing for customization based on the characteristics and needs of different games. In other words, DLSS 5 is not a "one-size-fits-all" technology; it has the potential to enhance game visual quality without disrupting the original style of the game.
So, let's mainly talk about the "AI feel."
Of course, we won't engage in professional and rigorous technical discussions here, and most people won't pay attention to the technical details behind DLSS 5. But they can all see that the "cost" of using DLSS 5 for visual upgrades objectively exists, which is the so-called "AI feel"—this is also a common adjective many people use to describe images generated by DLSS 5.
To discuss the "AI feel," we must first define what the "AI feel" is. However, this concept is too abstract and subjective, with everyone having different opinions, making it difficult to summarize a unified standard. Here, I can only share my subjective judgment—"high-density information" and "strong atmospheric lighting." In the earlier stages of generative AI development, specifically when image-generating AI was rising, I would also add "perfect composition."
Due to space constraints, I won't go into detailed explanations here, but I believe friends who frequently use AI to generate videos or images should know what this "feeling" is all about, just like in the image below:
Currently, the effects of DLSS 5 focus on "lighting shaping" and "detail rendering."
Moreover, perhaps due to the training model, its approach to shaping character lighting is quite "crude," looking as if a fill light was placed in front of the character, followed by cranking up the contrast and sharpening parameters, ultimately achieving results of "high-density information" and "strong atmospheric lighting." This is a major reason for the "AI feel."
In areas where this effect is particularly prominent, such as the wrinkled face of the old woman in Hogwarts Legacy, the "AI feel" is also stronger than in other places.
Here, I'll share another set of comparison images focusing on Grace's face.
Since Grace's face is relatively smooth (with lower information density) and the lighting atmosphere is similar, with only differences in the level of detail, the effect of DLSS 5 in this scene is closer to a "visual upgrade" rather than a "head transplant surgery."
The examples above may illustrate that the "AI feel" is not a preconceived error on the part of the audience—the leap in image quality is real, but consequently, the change in style and severe "AI feel" in some images are also real.
Or rather, at the current stage where DLSS 5 technology is not yet mature, they are inherently intertwined.
But considering that DLSS 5 is still an experimental technology, whether you want to resist it or embrace it, you'll have to wait for some time. After all, the current visual effects require two RTX 5090 graphics cards to achieve. Even by this autumn, there will only be demo versions running on a single RTX 5090 card, which simply cannot "reach ordinary households." If we want to popularize this technology on a large scale, we might have to wait until the 60-series graphics cards are equipped with neural computing units.
Considering that Moore's Law has failed and GPU computational power improvements are slowing down, perhaps DLSS 5 is indeed a clear path forward.
Although the current performance of DLSS 5 is less than satisfactory, advancements in AI technology always surpass people's expectations. The "Smith Eating Noodles" video AI incident is still fresh in our minds, but who could have imagined that within just a few years, they would develop to the point of being indistinguishable from reality?
Perhaps, with continuous technological iterations and gradual hardware upgrades, the "AI feel" in game visuals using DLSS 5 technology will significantly decrease, reaching a level acceptable to the general public.
