Last night NVIDIA announced DLSS 5, what they are calling the company’s most
significant breakthrough in computer graphics since real-time ray tracing and well there’s been a mixed reaction.
According to NVIDIA, DLSS 5 takes the colours and motion vectors of each frame before then feeding it into their AI model to create photo-real and lighting and materials. Now the use of gen AI here of course is going to immediately get people up in arms. But I don’t think the only issue is the idea of AI being used. I think the more we chase realism, the more uncanny and jarring the jump will be. Especially when we’re going back and retroactively applying it to games that we already have an established perception of the characters. It really doesn’t help that the model that they’re using does look like the average AI generated video.
There were a few frames or games where it did look like an improvement. Starfield for example, I think might have actually been an improvement but then when I’m looking at the comparisons for Resident Evil Requiem and Hogwarts Legacy, it feels like a downgrade. I spent way too long on NVIDIA’s 4k comparison of Grace trying to pinpoint what part of my brain doesn’t like it. The lighting while seemingly an improvement, makes the rainy Wrenwood street much less moody, Grace looks like a completely different character, darker features are overly prominent, her lips seemingly become fuller and she just generally looks older.
Some of this could just come down to the frames they have given us combined with the photorealistic lighting as some other clips I have seen look slightly better but something is being lost here. I do think already having an established idea of how the character looks beforehand really doesn’t help though.
Now while I still have reservations about how this looks, NVIDIA do say that DLSS 5 provides game developers with detailed controls, so they can maintain each game’s unique aesthetic. Which I’m not really seeing right now in the Resident Evil example at least but perhaps it needs more tweaking as it won’t be available for the game until fall.
Of course the technology will improve with time and a lot of the big publishers already seem to be on board with this so regardless of if people want it or not it’s on the way. However it is seemingly only going to be available for the RTX 50 series cards right now, which does make me wonder how much time is worth spending tweaking it to ensure it maintains the game aesthetic, when most people aren’t going to be experiencing it just yet.
So I do think there are some genuine concerns with hardware costs rising and how much style might be lost in games that don’t take the time to properly tune it, photorealism works for some games but it’s not something I’d want every game to chase. But it is for now at least going to be an option we can turn on or off and from the reception I’ve been seeing it’s one that a lot of people are going to be playing without.
Leave a Reply