• Hey Guest. Check out your NeoGAF Wrapped 2025 results here!

Nvidia at Live GTC : DLSS 5

Surely the strength of this effect will be customizable. I imagine a very light pass of this will be great for improving subtle details and lighting without dramatically changing faces or the overall art direction. The examples shown are so heavy-handed, though. But some people might like them, so hell, give them the option to go full AI nonsense.
 
I'm a massive fan of cutting edge graphics tech but what DLSS 5 seems to be doing is slapping one of those reshade photorealism mods on everything and ruining the game not to mention butchering the characters into those AI slop puppets. Oh yeah, I so want to pay 5K on a GPU so I can have every game I play ruined by that shite. Games are already suffering from looking samey, this shite makes all the games shown have the same drab "photorealistic" lighting.
If you're into that more power to you, I'm just glad I still have a massive backlog because if all new games converge into a stew of whatever the hell nVIDIA just showed off, they can keep it.
 

Danger 5 Laughing GIF

The meme became real
 
In Starfield example it looks like it works similar to Screen Space Global Illumination, with obvious issues when it's occluded by something.

I'm not sure about this tech in its current form...

yeah. the NPC moving in front of the bench completely fucked up the shadowed area below it.

and the fact that they need 2 5090s due to the sheer amount of memory this simple screen space based effect uses, makes it also unlikely that they can broaden what the ML algorithm can actually see without absolutely exploding memory usage even beyond this.

machine learning is insanely memory hungry
 
It's clearly not "just lighting"; nVidia themselves also included "materials" in that statement.
Ok I found the doc.

The AI model is trained end to end to understand complex scene semantics such as characters, hair, fabric and translucent skin, along with environmental lighting conditions like front-lit, back-lit or overcast — all by analyzing a single frame. DLSS 5 then uses its deep understanding to generate visually precise images that handle complex elements such as subsurface scattering on skin, the delicate sheen of fabric and light-material interactions on hair, all while retaining the structure and semantics of the original scene.

DLSS 5 provides game developers with detailed controls for intensity, color grading and masking, so artists can determine where and how enhancements are applied to maintain each game's unique aesthetic. Integration is seamless, using the same NVIDIA Streamline framework used by existing DLSS and NVIDIA Reflex technologies.
 
Some of these can't be real - just looks like a really bad filter app or something. They change the original artistic intent way too much.

I think maybe the idea is that developers can use these tools to create more realistic images - but can tailor it more specifically to what they want? Otherwise I don't know.
 
DLSSlop 5
Looks "good", I guess. Just like AI powered mods and reshade. Way to murder game art



Maybe it is a psyop by the big publishers to convince us that AI generated graphics look "good" so they can replace all art with AI slop
Mmmm yes the stunning game art of starfield etc ?
 
DLSSlop 5
Looks "good", I guess. Just like AI powered mods and reshade. Way to murder game art



Maybe it is a psyop by the big publishers to convince us that AI generated graphics look "good" so they can replace all art with AI slop
It's a turn off for me. With or with an AI slop filter they're gonna reach this real life version eventually so we probably should try get used to it.

Personally I like very realistic games but I also like the gaming art style that provides an escape from the real world.
 
Yep.

-Devs do face scans to try to make the character the most realistic possible... looks like mannequin in the end.
-Some players: Looks incredible!!!
-This tries to make it literally photo realistic.
-Same players: looks like garbage/AI slop.

Gamers are very annoying. They complain that the company doesn't pay attention to them, saying that they've abandoned gamers.
When the company demonstrates a new technology that is OPTIONAL, they complain.

I understand the point that there's a lot of garbage made with AI out there. But the level of reaction from people is bordering on mental retardation.

Wrong.



See : Materials

Material Depth: Enhances PBR properties like roughness and adds micro-realism to complex objects such as eyes and hair.

Simply apply new lighting to the existing material.

 
Cool stuff. I'm sure the devs will continue to have all the artistic authorship they need so I wouldn't worry about such things. Besides, graphics mods exist, and who's to say what the correct way to experience games is anyways? Is the correct look in RE:Requiem for Grace the one without RT, or with PT?

I bet the "AI Slop" opinions will suddenly change for many once AMD, Sony or Nintendo ends up doing their version of this stuff.

The dual 5090 does beg the question though how performance intensive this will be in the end. I wouldn't actually mind if Nvidia brought back SLI support if it meant better DLSS/frame gen performance like with Lossless Scaling. I do have a secondary 3060 I could use in addition to my 4090, but I bet they'd only support NVFP4 hardware if they did that.
 
Ok I found the doc.

The AI model is trained end to end to understand complex scene semantics such as characters, hair, fabric and translucent skin, along with environmental lighting conditions like front-lit, back-lit or overcast — all by analyzing a single frame. DLSS 5 then uses its deep understanding to generate visually precise images that handle complex elements such as subsurface scattering on skin, the delicate sheen of fabric and light-material interactions on hair, all while retaining the structure and semantics of the original scene.

DLSS 5 provides game developers with detailed controls for intensity, color grading and masking, so artists can determine where and how enhancements are applied to maintain each game's unique aesthetic. Integration is seamless, using the same NVIDIA Streamline framework used by existing DLSS and NVIDIA Reflex technologies.
I think a lot of the complaints are from it attempting to overdo the changes to the skin materials and reflections on them. It just looks too shiny, which is how a lot of GenAI stuff looks. Would like to see the RE stuff with the intensity turned down a bit.

The girl from RE also looks so different because of the shadowing on her cheek which is largely just lighting changes but she also looks like she just put on a ton of makeup or something.
 
Last edited:
Some of these can't be real - just looks like a really bad filter app or something. They change the original artistic intent way too much.

I think maybe the idea is that developers can use these tools to create more realistic images - but can tailor it more specifically to what they want? Otherwise I don't know.
But what if the developers chose how the filter was applied - then it becomes the artistic intent. People turn off motion blur, vignetting, chromatic abberation, film grain all the time - even if the original artistic intent was to have them applied.
 
It's simply spectacular (next-gen, no doubt about it) but I see two issues. 1. It looks like DLSS combined with Unreal Engine is going to give it a generic, AI-generated look, which I don't like one bit. 2. How many thousands, and thousands and thousands of dollars will it cost to build that PC with its mini nuclear power ☢️ plant just to play *Starfield* with PS5-level graphics?
 
Last edited:
If NVIDIA changes how the games are made by their makers, what's the point of this shit??
This:

DLSS 5 honors artistic intent in two ways:

  • Inputting the game's color and motion vectors for each frame into the model, anchoring the output in the source 3D content.
  • By providing developers with detailed controls such as intensity and color grading. Artists can use these controls to adjust blending, contrast, saturation, and gamma, and determine where and how enhancements are applied to maintain the game's unique aesthetic. Developers can also mask specific objects or areas to be excluded from enhancement.
 
When solely going from the images posted here....yes that does look like the AI stuff you see everywhere nowadays. But the DF video really shows parts of RE that make it look really good. To be honest I never understood why the Grace character model's face looks so simplistic compared to Leon's, and just in general in that street part.

Some of this looks quite good, other parts just look like it totally strips it away from what the creator intended. I'm personally not very impressed with the Oblivion parts.
 
And a beautiful part is that you can turn it off or the developers can make different uses out of it. It's no point in inferring what they can do when it will be up to developers and how they use this. I'm not sure what the anger is being placed at but I think it's just shooting at anything when people that are being cynical don't really know the target they are aiming at and it's actually pretty pathetic to read some of the statements. You appreciate but some people are just being goofy and that's not really what this threat should be about as much as theorizing and talking about the tech and not necessarily mocking it with low hanging fruit insults.
 
Last edited:
To be honest I never understood why the Grace character model's face looks so simplistic compared to Leon's, and just in general in that street part.
To make her kind of look like an anime character would be my guess. Unrealistically smooth faces are kind of the norm for that.
 
Top Bottom