DLSS 5 will be the worst thing to EVER happen to video games

Tea Cozy · 2026-03-17T20:50:59+0000

TwoSix said:
If your belief is that the designers who worked on the left-hand image didn't wish they had the tools to create the right-side image with real-time rendering, my gut feeling is that you're mistaken.

Capcom and Bethesda seem to agree (at least officially) with you.

Todd Howard of Bethesda:

“NVIDIA and Bethesda have a long history of pushing gaming graphics and innovation forward, and DLSS 5 represents the next major step in that journey," said Todd Howard, studio head and executive producer at Bethesda Game Studios. "With DLSS 5, the artistic style and detail shine through without being held back by the traditional limits of real-time rendering. We’re excited to work with this new technology and look to bring DLSS 5 to Starfield and future Bethesda titles.”

Jun Takeuchi of Capcom:

“At CAPCOM, we strive to create experiences that feel cinematic, compelling and deeply believable — where every shadow, texture and ray of light is crafted with intention to enhance atmosphere and emotional impact,” said Jun Takeuchi, executive producer and executive corporate officer at CAPCOM. “DLSS 5 represents another important step in pushing visual fidelity forward, helping players become even more immersed in the world of Resident Evil.”

Source: NVIDIA DLSS 5 Delivers AI-Powered Breakthrough in Visual Fidelity for Games

I’m personally one of those who think the images on the left are too smooth (the Resident Evil one looks especially doll-like) and the images on the right introduce much-needed detail.

As for preserving the intended lighting mood or overall art direction, I think that’s where the developer side of DLSS comes in.

Aeson · 2026-03-17T21:57:27+0000

Another AI is evil whaaaa

thread. Yippee.

Voranzovin · 2026-03-17T22:19:15+0000

As someone who works with real-time rendering professionally, the concern here isn't whether the result is better or worse, but whether it's more generic. It subtly changes the character's expression, potentially erasing artistic choices. This particular character has a fairly nondescript face to begin with, presumably chosen because it makes a good demo. How well it will work on characters with stronger art direction remains to be seen.

That said, this isn't the kind of AI use I personally object to on principle; this looks like an advanced form of style transfer, which is a technique that's been used in film and game production in various ways for years, back when we called this kind of thing "machine learning" and no one thought anything of it. I have my doubts about whether it will prove to be a useful tool for anything but the most generic photorealism, but I wouldn't lump it in with the twisted mockery of life conjured by Midjourney and its dark spawn.

Morrus · 2026-03-17T23:02:31+0000

Aeson said:
Another AI is evil whaaaa thread. Yippee.

Another critics of AI are evil whaaaa

post. Yippee.

See how easy it is to just mock people who disagree with you? We expect better than that here, please.

bsss · 2026-03-17T23:11:17+0000

I'm a pretty heavy skeptic towards generative AI, especially when misused to replace human output, but yeah, my original thought on this was "I remember when this was known as "applying neural networks" or other machine learning, and people were impressed". A lot of the outrage coming hot on DLSS 5, IMO, is a feedback loop of dunking on the "yassified" "instagram filter" examples plus lumping in all "AI" in with "slop". Meanwhile, there are some other examples in the demo footage that look like pretty impressive lighting changes to be applied in realtime. Does the RE9 example show a heavily modified face, arguably for the worse (from an artistic choice perspective, at least)? Yes, sure. Maybe it's a bad example, maybe the model is just not great with faces in its demo stage. Maybe it'll always be crap at faces. If it turns out to be not worth it, life will move on, developers won't bother with it, Nvidia will come up with a DLSS 6 or whatever. Deep learning is a valid technique quite different from "the slop".

gban007 · 2026-03-18T00:59:56+0000

bsss said:
I'm a pretty heavy skeptic towards generative AI, especially when misused to replace human output, but yeah, my original thought on this was "I remember when this was known as "applying neural networks" or other machine learning, and people were impressed". A lot of the outrage coming hot on DLSS 5, IMO, is a feedback loop of dunking on the "yassified" "instagram filter" examples plus lumping in all "AI" in with "slop". Meanwhile, there are some other examples in the demo footage that look like pretty impressive lighting changes to be applied in realtime. Does the RE9 example show a heavily modified face, arguably for the worse (from an artistic choice perspective, at least)? Yes, sure. Maybe it's a bad example, maybe the model is just not great with faces in its demo stage. Maybe it'll always be crap at faces. If it turns out to be not worth it, life will move on, developers won't bother with it, Nvidia will come up with a DLSS 6 or whatever. Deep learning is a valid technique quite different from "the slop".

Yeah, I think im in similar boat. I think applying it to existing work / games can cause issues, with characters looking different pre and post , with some looking like completely different people which could be disconcerting.
However, the general quality does look better, and so I think for new projects/ games this could be quite beneficial, if hard on machines.

nyvinter · 2026-03-18T01:05:33+0000

It's probably great if you want every character to look liker they have a ring light in their face at all time

I don't think Starfield looks great normally, but that hanger at least has shadows and internal lightning that makes sense. The DLSS 5 removes all of that for a harsh noon sunlight, even indoors. And in the Resident Evil it removes the background fog. Can't have mood and atmosphere.