• 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

might be fake?​

​

“DLSS 5 Gets Modded Into Skyrim With Dramatic Changes to Lighting and Character Faces​

by DrUsmanSaleem Today, 04:53 Discuss (4 Comments)
NVIDIA announced DLSS 5 back in March to a decidedly mixed reaction, as many gamers did not particularly like the modifications it made to character faces. While it is not officially out yet, an early-access build of NBA 2K27 was found to contain a DLSS Neural Rendering DLL, believed to be a precursor to DLSS 5. As expected, modders have been quick to start implementing that Neural Rendering technique in games, and what better game to add a new mod to than The Elder Scrolls V: Skyrim.

YouTuber PureDark implemented the DLSS-NR mod into Skyrim and recorded the results. Skyrim, now 15 years old, has some pretty outdated character models, but this early DLSS 5 implementation really brought those faces into 2026. In these side-by-side shots, a dramatic difference is evident between the two sets of character faces, with superior detailing, lighting, shadows, and even facial structure. The overall lighting behavior of the game is vastly improved too, and it almost looks path-traced in some scenes“
 
Last edited:
M76 said:
True 4K or upscaled 4K? Very important distinction. If it's true 4K that's pretty good, but I doubt it.
Click to expand...
I think the new neural part of DLSS 5 is applied at the end, so I am not sure the distinction matter, for the neural part that the 4k image was native or upscaled it could be the same cost.... (as long that they made a dlss 4.5->5 at the same scaling, which I assume they did here)

The fact that it is being ran on 4k of pixels is a lot.

I imagine a next evolution could be a "one-go" where the upscaling and neural construction is made in one go/same model
 
Last edited:
LukeTbk said:
I think the new neural part of DLSS 5 is applied at the end, so I am not sure the distinction matter, for the neural part cost that the 4k image was native or upscaled it could be the same cost.... (as long that they made a dlss 4.5->5 at the same scaling, which I assume they did here)

The fact that it is being ran on 4k of pixels is a lot.

I imagine a next evolution could be a "one-go" where the upscaling and neural construction is made in one go/same model
Click to expand...
i just don’t like this

feels gross for some unknown reason

i’m with the people playing games to escape and relax outside of reality not trying to simulate photorealism

wasn’t there a photorealism mega-thread on here somewhere?

ah well, to me this feels like ai slop
 
erek said:

might be fake?​

Click to expand...
Seem that would be a lot of work to generate a fake video....

This is quite impressive for an absolute worse case, the particle and moving leaf seem well handled but that could be because the model mostly look at characther, it could very quickly become a total non-brainer to use...

erek said:
ah well, to me this feels like ai slop
Click to expand...
Quickly the other way around, when it is set to off it is like some strange artifical botox filter is applied to the whole image, that remove contrast and flaten the light.
 
erek said:
trying to simulate photorealism
Click to expand...

That's been the goal since 'the first pixel' (so to speak, was before even pixels).

It's the reason for every tech advancement, from 8bit to 16bit to 32bit to 64bit etc.

From sprites to textures.

Virtual worlds and simulations are a reflection of our own, the goal is and will always be 1:1 replication.

You can have your artistically styled games still, but tech doesn't advance with the goal to make things less accurate than reality.
 
Last edited:
LukeTbk said:
I think the new neural part of DLSS 5 is applied at the end, so I am not sure the distinction matter, for the neural part cost that the 4k image was native or upscaled it could be the same cost.... (as long that they made a dlss 4.5->5 at the same scaling, which I assume they did here)

The fact that it is being ran on 4k of pixels is a lot.

I imagine a next evolution could be a "one-go" where the upscaling and neural construction is made in one go/same model
Click to expand...
The distinction matters because if it's true 4K rendering then there is headroom to improve performance. If it's already running DLSS Balanced or Performance then it is practically unusable at 4K.
 
M76 said:
The distinction matters because if it's true 4K rendering then there is headroom to improve performance. If it's already running DLSS Balanced or Performance then it is practically unusable at 4K.
Click to expand...

Yes, more data is always better/garbage in, garbage out

But, you could always (also) lower settings 😱

And I'm interested to see where/what the graphical trade off is, is it better to lower settings (up to whatever point?) and keep neural rendering on, or turn it off for more enabled 'classical' graphics settings?
 
chameleoneel said:
It all just looks like the cover of a body building magazine.

Emphasize every contour/add specular highlights. Boost contrast of shadows and mid-tones. desaturate color.

In a movie, a DP would have to be extremely specific about lighting, to get people's Philtrum to look boldly contoured like that. And it wouldn't look more/less static like that, as they moved around the scene.

On the lowest "0" settings, its not necessarily egregious. But, the color is still heavily altered and desaturated (And Control already isn't very colorful!). And a lot of the intentional colored light bleed/global illumination, is sucked out.

Even giving this my most generous take--------not worth literally cutting FPS in half.
Click to expand...
I mean I agree but the concept of it has value IMO. Just like DLSS was scuffed at the start, this is going to be scuffed but I think it has good potential for the future
 
kram182 said:
That's been the goal since 'the first pixel' (so to speak, was before even pixels).

It's the reason for every tech advancement, from 8bit to 16bit to 32bit to 64bit etc.

From sprites to textures.

Virtual worlds and simulations are a reflection of our own, the goal is and will always be 1:1 replication.

You can have your artistically styled games still, but tech doesn't advance with the goal to make things less accurate than reality.
Click to expand...
i heard the less photorealistic games age much better

mario 64 isn’t trying to be photo real, or is one of the best games of all time zelda - ocarina of time

masterpieces

a photorealistic battlefield or CoD is likely to never be a masterpiece class game
 
erek said:
i heard the less photorealistic games age much better

mario 64 isn’t trying to be photo real, or is one of the best games of all time zelda - ocarina of time

masterpieces

a photorealistic battlefield or CoD is likely to never be a masterpiece class game
Click to expand...
What if you could get something like this?


View: https://www.youtube.com/watch?v=aL1GoWnChHc

edit: Before anyone points it out I know about the misspelling.
 
erek said:
i heard the less photorealistic games age much better

mario 64 isn’t trying to be photo real, or is one of the best games of all time zelda - ocarina of time

masterpieces

a photorealistic battlefield or CoD is likely to never be a masterpiece class game
Click to expand...

Now compare Super Mario Bros to Super Mario World to Super Mario 64 to Super Mario Galaxy to Super Mario Odyssey.....

Even Mario has and will continue to get more photorealistic as time and technology progresses, even if never on the cutting edge of and still stylized
 
kram182 said:
Now compare Super Mario Bros to Super Mario World to Super Mario 64 to Super Mario Galaxy to Super Mario Odyssey.....

Even Mario has and will continue to get more photorealistic as time and technology progresses, even if never on the cutting edge of and still stylized
Click to expand...
why not just jump to it then

doesn’t make sense

there were more photorealistic capabilities at the same time
rinaldo00 said:
What if you could get something like this?


View: https://www.youtube.com/watch?v=aL1GoWnChHc

edit: Before anyone points it out I know about the misspelling.
Click to expand...

it doesn’t feel good for some reason

i’m allowing my personal taste about it all

i don’t like it, and seen similar results a few years ago before i quit the gta 5 modding scene

it may of been a little slower but they essentially piped into “chatgpt” and the output reminds me exactly of that with dlss5

i even posted those examples before

it’s got this same type of look and it gets old and everyone was doing it and nothing felt good anymore
 
erek said:
why not just jump to it then

doesn’t make sense

there were more photorealistic capabilities at the same time
Click to expand...

Why aren't you only and exclusively playing 8bit sprite based 2D games still?

How dare you play and enable the progress of 'soulless', more realistic, 3D and texture based games!!!!!!!

No one says you have to like it, but you can't tell others they can't like it all the same.

Thus arguing against its existence is stupid and hypocritical.
 
erek said:
why not just jump to it then
Click to expand...
For the same reason photorealistic movies are still watched and enjoyed....

For one an horror movie or a war video game, can provide a cheap version of the trill without the actual danger, an even easier/cheaper way than the fake horror house/roller coaster.

Some video game can benefit from more photo realistic graphic (you can still be in a very unrealistic world or very artificial light for astheatic reason just well calculated one, same as games of thrones/star wars and most movies that use spots, color grading and many tricks, photo quality but real different than reality and not just from fantastic set design but light creation-manipulation-lens), some could get too much realistic from their own good, some can decide to be 2d, cell-shade, anything.

This does not need to be used to go toward photorealism either, this is just an easier first step because of the giant training data / universal rules that would always apply which make sense for a third party tech coming from Nvidia, but a game studio could go for something fully different look wise instead, just different training set and rules.
 
kram182 said:
Why aren't you only and exclusively playing 8bit sprite based 2D games still?

How dare you play and enable the progress of 'soulless', more realistic, 3D and texture based games!!!!!!!

No one says you have to like it, but you can't tell others they can't like it all the same.

Thus arguing against its existence is stupid and hypocritical.
Click to expand...
why do you like it
 
kram182 said:
Why aren't you only and exclusively playing 8bit sprite based 2D games still?

How dare you play and enable the progress of 'soulless', more realistic, 3D and texture based games!!!!!!!

No one says you have to like it, but you can't tell others they can't like it all the same.

Thus arguing against it's existence is stupid and hypocritical.
Click to expand...
Honestly, I’m gonna learn how to use the GoDot engine and see what I can do with that.
1080p target render and Linux friendly.

I just can’t get the idea of putting out games on SD Cards that target the PI that contain their own OS
so you can play them like a console.

Maybe using PICore as the base.

Because why not?
 
LukeTbk said:
This does not need to be used to go toward photorealism either, this is just an easier first step because of the giant training data / universal rules that would always apply, but a game studio could go for something fully different look wise instead, just different training set and rules.
Click to expand...

Yeah this could/should be able to be used for cell shading etc too 🤔
 
my vision isn’t very good so it’s diminishing returns once the fidelity and so-called realism hits a certain point

so i can appreciate a good story and the gameplay

i don’t find myself pixel peeping as i even became more of an gamer

before all i cared about was max graphics but i was a poser and didn’t even grasp anything about input latency or anything
 
erek said:
why do you like it
Click to expand...

Because I like the 'attempt to recreate reality of it all' from a technological perspective

And when I'm shooting zombies in Resident Evil etc I want them to look realistic for better immersion and goriness etc according to my own personal tastes and preferences
 
erek said:
why do you like it
Click to expand...
A bit like an small laptop can infer meteo at 1/1,000,000 of the cost of a supercomputer needing to calculate to do it, this as the potential to punch a giant amount of very advanced graphic on old wafer with very little watts.

In SRAM (Groc-cerebras) NPU, this could very well for 4k frame be done in 1-2 ms, letting and engine run with basic ground truth at 500-1000 fps on cheap hardware.

And sram rendering could be cycle-exact, i.e. always every 2ms generate a new frame from the best ground truth available, without a single frame time move between 2 frame ever occuring, this could all go really fast by PS7 release time.
 
kram182 said:
Because I like the 'attempt to recreate reality of it all' from a technological perspective

And when I'm shooting zombies in Resident Evil etc I want them to look realistic for better immersion and goriness etc according to my own personal tastes and preferences
Click to expand...
but they can achieve photorealism right now

so why not just do it
 
erek said:
but they can achieve photorealism right now

so why not just do it
Click to expand...
alan wake 2 with dlss5 mod, i’d like to see

1787934690909.png
 
erek said:
but they can achieve photorealism right now

so why not just do it
Click to expand...
Not sure what you mean, yes render farm for SFX team do it, but not in real time yet, so they compromise for performance (much better to make a game run at 60 fps with compromise than take 2 minutes per frame to achieve Toy Story 5 or latest Chris Nolan movie level of render)
 
LukeTbk said:
Not sure what you mean, yes render farm for SFX team do it, but not in real time yet, so they compromise for performance (much better to make a game run at 60 fps with compromise than take 2 minutes per frame to achieve Toy Story 5 or latest Chris Nolan movie level of render)
Click to expand...
i bet they can do it in real time for sure

beyond the optimization techniques (compression, dithering, denoising, etc etc, dlss5) and employing some artistry about it i am sure it can be done

i think clever artistry can push it over the line honestly

doesn’t have to be purely mathematically achieved in my opinion

more than one way to skin a cat
 
erek said:
but they can achieve photorealism right now

so why not just do it
Click to expand...

1) if you've seen any of the recent Resident Evil games, they do

2) photorealism is not a set thing (until we have a 1:1 simulated-exactly-as-ours-is universe/game/game engine) -it's an ever moving benchmark enabled by and achievable through new and newer technologies

Games are a technological product, like it or not. Therefore 'good enough' will never be 'good enough' for the technology side/aspect. There will always be a better on the horizon and a strive for that better.
 
kram182 said:
1) if you've seen any of the recent Resident Evil games, they do

2) photorealism is not a set thing (until we have a 1:1 simulated-exactly-as-ours-is universe/game/game engine) -it's an ever moving benchmark enabled by and achievable through new and newer technologies

Games are a technological product, like it or not. Therefore 'good enough' will never be 'good enough' for the technology side/aspect. There will always be a better on the horizon and a strive for that better.
Click to expand...
taking photos and projecting them into 3D and or voxel displays and so forth

definitely think you’ll hit diminishing returns unless we get holographic immersion

in seeing quite a few folks noting the uncanny valley about it all
 
erek said:
taking photos and projecting them into 3D and or voxel displays and so forth

definitely think you’ll hit diminishing returns unless we get holographic immersion

in seeing quite a few folks noting the uncanny valley about it all
Click to expand...

1) Still does nothing for the lighting/physics aspect of it, all tied to things like color and perception in the real world which is why we also have PBR and raytracing/pathtracing etc

2) you just previously claimed you can't see well and don't pixel peep, so how would you know to be frank?

3) People claimed uncanny valley with Playstation 2 and Playstation 3 level graphics etc as well, yet here we are well beyond that and what was then an uncanny valley is quaint and dated today, nor did it hold us back from even more realistic graphics and the strive for yet even more still

https://www.neogaf.com/threads/ps3-and-xbox-360-the-uncanny-valley.120704/

Stand in front of the steam train (ironically and fittingly) known as technological progress and try to stop it all you might, call it 'not progress' or 'uncanny valley' or whatever, it's still just gonna run you over and keep progressing/moving forward all the same.
 
Last edited:
kram182 said:
1) Still does nothing for the lighting/physics aspect of it, all tied to things like color and perception in the real world which is why we also have PBR and raytracing/pathtracing etc

2) you just previously claimed you can't see well and don't pixel peep, so how would you know to be frank?

3) People claimed uncanny valley with Playstation 2 and Playstation 3 level graphics etc as well, yet here we are well beyond that and what was then an uncanny valley is quaint and dated today, nor did it hold us back from even more realistic graphics and the strive for yet even more still

https://www.neogaf.com/threads/ps3-and-xbox-360-the-uncanny-valley.120704/

Stand in front of the steam train (ironically and fittingly) known as technological progress and try to stop it all you might, call it 'not progress' or 'uncanny valley' or whatever, it's still just gonna run you over and keep progressing/moving forward all the same.
Click to expand...
hmm i’ve seen retrospectives that argue we’ve regressed in graphical impressiveness in some aspects and plateaud for generations in others

i feel this is somewhat true

i still believe that photorealism is achievable in realtime with enough effort and creativity


i really do
 
kram182 said:
1) Still does nothing for the lighting/physics aspect of it, all tied to things like color and perception in the real world which is why we also have PBR and raytracing/pathtracing etc

2) you just previously claimed you can't see well and don't pixel peep, so how would you know to be frank?

3) People claimed uncanny valley with Playstation 2 and Playstation 3 level graphics etc as well, yet here we are well beyond that and what was then an uncanny valley is quaint and dated today, nor did it hold us back from even more realistic graphics and the strive for yet even more still
Click to expand...
Correct me if I am wrong but I thought that the uncanny valley gets worse the closer you get to almost human but not human. If true then it should be worse now. A 2D stick figure will not produce the effect but a 3D almost human figure would.
 
You must log in or register to reply here.
Back
Top