NVIDIA: DLSS 5

Photo by Nvidia - © 2026 All Rights Reserved.

In recent years, it could be said that technologically we are slowing down and heading towards a bit of a plateau in terms of the raw power increase. Think of it this way: how much better has the iPhone actually gotten in the last 3 years? Not the same exponential or even persistent gain as it used to be, it is now incremental. Video games have similarly suffered the same fate. The jump into the newest generation of consoles and graphics cards has not been as substantial as it was during the mid-2010s. Whilst the price of hardware and video game development has been steadily going up, it has certainly not yielded the results that certain people had hoped for. There are cases of great success, yet ultimately, people are not really impressed. Visual fidelity improvements in games seem to be a tricky thing to handle. That is not something new for this generation; since their dawn, video games have used very clever tricks to make the impossible happen, even with their limited technologies for their times. Games came out sometimes looking like they were made in the future. Yet that era had a huge exponential increase in raw technological power. Right now, we are slowly starting to approach our limits based on physics, such as transistor scaling and power and heat management, to name some. In addition to that, advanced game development tools like Unreal Engine 5, for example, boasted promises of future graphical advancements. However, a vast number of games developed on it look mediocre at best. Even if they look good, most of the time their performance makes you question the price that you have paid for this "Next Gen Experience". More power = more possibilities, that was the vision. A vision that is now reaching a technological wall that many companies have tried to solve using software enhancement. 

​

NVIDIA, being the giant that they are, saw this wall coming, as pre-2020, they decided that the helping hand that would allow us to get over this wall is, unsurprisingly, that of AI. In this case, think of AI as a force multiplier when it comes to the raw technological limit. DLSS 1 (Deep Learning Super Sampling) was initially meant to natively render the game at a lower resolution to increase frame rate performance and then upscale the image using AI. Though faulty at first, each iteration grew to be widely implemented in almost all games today. The next breakthrough came with DLSS 3 (2023), where AI would not just upscale, but generate "fake" (not natively rendered) frames in between the native ones. Again, faulty at first, but now quite well accepted amongst gamers. Each iteration used AI to help boost the game's visual performance when our hardware hit the limits, but up until 2 days ago, it never actually overhauled visuals fully. NVIDIA comes in with a potential breakthrough in the path they set with the announcement of DLSS 5. At its core, what this new AI does is it infuses the pixels with new photorealistic lighting and materials, see examples here. The result... is quite game-changing. The internet's response to this: absolute dumpster fire. To sum it up, the majority of people are simply calling it: "AI slop," and potentially rightfully so. I am a bit torn on this, to be fully honest. I do see the potential that this technology is bringing, yet the concern of artistic integrity and vision being overhauled by AI is very concerning here. NVIDIA claims that each game developer has full control over how to use DLSS 5 and underlines that they still have full artistic control. Yet, examining the examples that they provided, especially noticeable on the faces of the models, is the generic AI style of "make it hyper realistic" prompting. Making the characters look at times as if they were put through an Instagram filter.

 

 

 

 

 

 

 

 

 

 

 

 

What makes it stick out is the difference in how the model was originally rigged and designed to sell the illusion of realism or artistic intent that the game is going for. DLSS 5 puts a "prettier" filter over the pixels, but right now, it seems too generic and not properly adjusted to the original aesthetic design. Which produces an image bland like every other AI content you stumble upon on your social media these days. Notice that not a lot of gameplay or footage was shown. As even in the very quick snippets, when the faces move, something just seems off. It's the movement of the smaller facial features that are now more detailed in light and texture, but not in their original animation. You see, for example, what makes films like Avatar stand out, especially in photorealistic facial animation, is that they have a lot more motion detail from face capture than a video game model requires. Since a game has to render the image on the go, it needs to be efficient at allocating resources in proportion to its system's power capabilities. Therefore, when DLSS 5 reworks the image, it highlights areas that do not possess the recorded, smaller movement details in our face that our brains recognize as authentic. Ergo, the image looks off. 

 

The kicker is that all of the games that are showcased had to be approved by their developers, which means they were fine with this outcome. I believe that is what really upset the community. Making graphical improvements is always welcome, yet art cannot be made generic, especially today, when AI is more integrated into our lives than ever before, and it does not seem to be slowing down. There is potential in this technology, but it has to improve like its previous iterations, and it must aid, not overtake human artistry. The community backlash is understandable. Only time will tell how this unfolds, as right now, this technology is in a very primitive state and will be limited to high-end PC NVIDIA graphics cards, yet we all know copies and general implementation of it could very soon be seen.

​