• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

DLSS 5 - Generative AI

https://x.com/ryanshrout/status/2033686038829535318

1773715906136.png


(more written in the link)

Edit: "iT mAkEs EvErYoNe LoOk ThE sAmE uNLiKe BeFoRe Ai"

Meanwhile, 'gaming before AI':
1773716396354.png
 
Last edited:
What people don't get is that DLSS5 is not enhancing the image so much as filtering it. Which sounds cool until you realize that every game developer is pissed off because you're altering the look of the game. Not that we don't already do that in a myriad of different ways.
 
Sycraft said:
I think the real angy with this is because it is "AI" and people are just being angy about that without thinking (remember that DLSS upscaling is "AI" as well) and because it sounds like it is going to need a 50 series minimum, and probably a beefy one at that and there's sour grapes from the people who don't have that. "It's not that I can't run it, it is that I don't WANT to run it!"
Click to expand...
Those who follow image model progress know there are interesting ways they can be utilized and finetuned that don't draw negative attention to themselves. What they showed though looks like the kind of immediately identifiable AI img2img 'improvements' popularized by homogenized models and prompts for years.

Nvidia's response conspicuously doesn't mention there's a way for devs to adjust the style or provide something like a LoRA. Only its intensity, color grading and where it's isolated to. Given all the examples follow a similar adjustment it's not surprising people are skeptical whether the style can be changed.

Take say the Starfield example. The rest of the scene remains largely the same style (non-photorealistic), while the foreground face is conspicuous for not matching the graphics of everything else. That's one kind of disconnect. Or the Grace example with Requiem where the model seemingly mistakes the under-eye lighting for eyeshadow makeup, turns her eyebrows dark in their entirety instead of blonde and makes her irises darker.

There's a known ground truth for these characters: the higher fidelity cutscene versions and promotional images. In those cases it looks off when players know what they look like in high def and see the enhancements off in subtle ways. Like sure, ostensibly it may be more detailed but humans (outside of those with face blindness) are very good at picking up on things not quite right with faces (even when aiming for accuracy like CGI face doubles in films).

Edit: for things like environments I do think there's an interesting place for real-time models that could enhance low setting graphics or do style transfer. They just led with something that doesn't land for many.
 
Last edited:
Sycraft said:
3) It is OPTIONAL, just like frame gen, ray reconstruction, etc, etc. If a developer doesn't like it, they aren't required to use it.
Click to expand...
People know that it might be optional now but there's a good chance it won't be in the future. You forget Microsoft bought a whole lot of game development studios. You think they will allow it to be optional? Not likely. Microsoft in particular needs to recoup its Open AI investments and this is a good way to do it.
Sycraft said:
I think the real angy with this is because it is "AI" and people are just being angy about that without thinking (remember that DLSS upscaling is "AI" as well) and because it sounds like it is going to need a 50 series minimum, and probably a beefy one at that and there's sour grapes from the people who don't have that. "It's not that I can't run it, it is that I don't WANT to run it!"
Click to expand...
People didn't have a problem with it before because it was limited for the most part. The most intense stuff was limited to Remedy. Outside of that you might get some graphical anomalies here and there, but overall DLSS improved the image that was already there. This isn't that. This is now changing artistic intent on a much bigger scale and that's what people are pissed about. We went from filling out the inside of a chain link fence, to putting make-up on a supposedly disheveled character that game developers wanted to be insecure. That's the difference. The graphic artists in RE9 wanted Grace to be a freaking mess. Nvidia comes a long and makes her besties with Paris Hilton. In addition there's some changes to Grace that affect the story line. I'm not going to give it away in case people haven't played it but her appearance changed enough that it will change the story of the game.

Not a huge fan of MLID but this video with a game developer is good for it's perspective.


View: https://youtu.be/0Tiw7blVRZg?si=lgStsaTfyAQ7LTDJ&t=367
 
Last edited:
kac77 said:
The graphic artists in RE9 wanted Grace to be a freaking mess. Nvidia comes a long and makes her besties with Paris Hilton.
Click to expand...

They also intended her to look like a real person which is why they modeled her after a real person, that it looks more like now

Edit: also

1773726818560.png


1773722932873.png
1773722941998.png
 
Last edited:
At the end of the day, if the artistic intent of a game is designed with such an implementation in mind - meaning the results are 'as intended' by the game studio/developer, I don't think I have a problem with it.

I'm certainly not going to lay judgement until I see the results with my own eyes. I laid judgement on DLSS + FG before seeing both implementations, and now I use the features all the time where necessary - They're actually quite useful, and I struggle (with the latest implementations of DLSS I basically can't see a difference) to notice any decrease in visual clarity or increase in latency with Nvidia Reflex enabled.
 
IMHO, this is an extreme tease. I don't think we'll see this tech for years (?), at least not what they demo'd. Positioning. Yt'ers are saying none of what was demo'd was hinted at in any road map or futures materials. So... I think, it's a tease perhaps to get competitors focused on "bad things", etc.
 
cjcox said:
IMHO, this is an extreme tease. I don't think we'll see this tech for years (?), at least not what they demo'd. Positioning. Yt'ers are saying none of what was demo'd was hinted at in any road map or futures materials. So... I think, it's a tease perhaps to get competitors focused on "bad things", etc.
Click to expand...
They were stating relase a fall this year and people were having access to games running it now in real time so I would argue your claim is off base.
 
I can see this being useful if not applied in the overly aggressive way that they’re demonstrating it in, but it still needs work. Otherwise you’ll end up up with over processed ai slop.

IMG_0367.jpeg
 
kram182 said:
They also intended her to look like a real person which is why they modeled her after a real person, that it looks more like now

Edit: also

View attachment 791828
Click to expand...
That commnet show would I never listen to people like "Moore's Law Is Dead" as his bread and butter are hyping (mostly rumours) and now he jumped the bandwagon of ignorance like the majority of posters online.
 
kac77 said:
People know that it might be optional now but there's a good chance it won't be in the future. You forget Microsoft bought a whole lot of game development studios. You think they will allow it to be optional? Not likely. Microsoft in particular needs to recoup its Open AI investments and this is a good way to do it.
Click to expand...
I mean, The Xbox can't do it so that alone is going to make it optional for them. Also this is not something that helps with investment in something like Open AI. This runs local, on a GPU, in realtime. It's not a case of "Oh you have to use our cloud to play your game, requires too much power to run local!" This is local neural networks, just like the rest of DLSS. That's kinda the opposite of what the people who have big datacenters running teh stuff want. When

kac77 said:
The graphic artists in RE9 wanted Grace to be a freaking mess.
Click to expand...
So then is the Path Traced Grace fucked up? I haven't played the game yet but looking at the videos there's a HUGE difference between no-RT Grace that you get on the Xbox or base PS5 and Path Traced Grace that you get on the PC (with the RT being somewhere in the middle). Is the PT version wrong? There's a picture further up in the thread and she looks like more of a mess... but less real. In particular her eyes look stranger, they don't have good lighting and shadowing on them and her hair looks stringier and more fake because, well, it is it isn't being rendered with the same fidelity. So, is the PT one wrong? Did Capcom do themselves a dirty by putting it in the game? Because Path Trace Grace looks more like a real human

To me, DLSS 5 looks like another step beyond that. Now the self-shadowing makes her eyes look really three dimensional and set with the features of her face, her hair looks more correct for a dye-job, her jacket materials look more correct, etc. It's not perfect, but it does seem like another step towards more realistic lighting.

kac77 said:
Not a huge fan of MLID but this video with a game developer is good for it's perspective.
Click to expand...
You know, I'm just not going to get that worked up about it. We'll see how it looks when it is released, we'll see how it compares in game both performance and visuals wise.

I just feel like this is all people getting worked up mostly because of the "AI" tag. "Oh shit, it looks different and they said AI, therefore I must hate it!"
 
There’s also the possibility that nvidia is prepping for the bottom falling out of the AI market eventually. Then they’ll pivot to trying to sell us SLI again.
 
I am torn on this. If the game already has good graphics and art direction, the end result is obviously very negative, uncanny and awful.

But take a game that has very outdated graphics like the Starfield example here, or some older game like any Xbox One or PS4 era take on "realistic graphics". It just might be an improvement because the graphics were not very realistic in the first place. I would like to see what this does in Witcher 3 Next Gen to be sure. It probably mangles the looks of the game but who knows.
 
With the way things are going, why would this matter? The likelihood it will run on current mainstream or even performance GPUs is slim to none. New GPUs aren't coming out anytime soon and even when they do come out there won't be any real volume because everything is going to support the AI bubble.

Until the AI bubble pops this is effectively useless. It will probably still stay effectively useless until a couple more GPU generations come around. The fact that they used dual 5090s to run this demo to ensure it didn't screw up doesn't bode well for future use until there's a considerable performance increase which makes its way down to at least mainstream GPUs. It's likely to be even worse considering all the other things running at the same time on the GPU. There's only so much horsepower to go around and that is much worse on anything not in the halo model range.
 
SmokeRngs said:
With the way things are going, why would this matter? The likelihood it will run on current mainstream or even performance GPUs is slim to none. New GPUs aren't coming out anytime soon and even when they do come out there won't be any real volume because everything is going to support the AI bubble.

Until the AI bubble pops this is effectively useless. It will probably still stay effectively useless until a couple more GPU generations come around. The fact that they used dual 5090s to run this demo to ensure it didn't screw up doesn't bode well for future use until there's a considerable performance increase which makes its way down to at least mainstream GPUs. It's likely to be even worse considering all the other things running at the same time on the GPU. There's only so much horsepower to go around and that is much worse on anything not in the halo model range.
Click to expand...
This is a strategic move from NVIDIA just as they annouced DLSS a short time before the current consoles and thus kept a visual advantage through this whole console generation.
Now there are talks about the "next gen" consoles and their hardware is basically determined and their software features planned and Now comes DLSS 5 and jumps the rendering technology 1-2 generation ahead ensuring that all the companies trying to unify and make the consoles dictate the progress have been outplayed.

They will now have to spend another generation trying to catch up to NVIDIA, while NVIDIA can prepare their next move while refining DLSS5 at the same time.
There is a reason why their market share the last quarter was 94% because everyone else is chasing their tail trying to catch up to them and when they think they have, NVIDIA pulls out their next card and the cycle repeats.
 
DukenukemX said:
What people don't get is that DLSS5 is not enhancing the image so much as filtering it. Which sounds cool until you realize that every game developer is pissed off because you're altering the look of the game. Not that we don't already do that in a myriad of different ways.
Click to expand...
It feels like a reshade filter doesn't it.
 
MaZa said:
I am torn on this. If the game already has good graphics and art direction, the end result is obviously very negative, uncanny and awful.

But take a game that has very outdated graphics like the Starfield example here, or some older game like any Xbox One or PS4 era take on "realistic graphics". It just might be an improvement because the graphics were not very realistic in the first place. I would like to see what this does in Witcher 3 Next Gen to be sure. It probably mangles the looks of the game but who knows.
Click to expand...
Maybe they will add prompting to DLSS 5.1 :)
"Light witcher 3 like the witcher netflix show" lol
I am only half joking.

They may be trying to sell game publishers on the "you will have control" crap. I doubt that lasts long at all. Game publishers, the big ones anyway will jump in and say glowing things cause everyone has investors and you can't hate on AI while everyone is surfing the top of the bubble. I don't expect it will be long before Nvidia adds drop down boxes (or actual prompt options) for various light "art" styles. I mean why not... why not render that hello kitty game all dark and moody, or that dark and moody game all happy and sparkly. You want your Star Trek game to have JJ style lens flair... I mean the AI can clearly add it. You want it removed, can do that too.
 
You must log in or register to reply here.
Back
Top