Support NeoGAF

[kevboard]

bullshit ? stop spreading fuds? Nvidia them selves said the artist can even control and adjust the enhanced image to their liking. this is not a random chatgpt garbage.

I am glad this tech exists. so it pissess off people like you even more. go buy AMD product...

another one who doesn't know how generative AI works.

again, what I said before still stands tech illiterate people love it because they don't know how it works.

it is LITERALLY impossible to look the same each time. if you turn the camera away from a scene and then back again, you will see differences, even if all conditions are identical. same camera view, no lighting changes, no dynamic objects being moved... it will never look the same.

in order for generative AI to look similar each time (note I say similar, not the same, because... repeat after me: IT'S LITERALLY IMPOSSIBLE to look the same) it will need to remember how it looked the last time you saw that exact scene. otherwise there's no guarantee it will even have the same shadowing or color temperature, especially if vital information about the lighting conditions aren't fully evident inside screen space.

 

[//DEVIL//]

another one who doesn't know how generative AI works.

again, what I said before still stands tech illiterate people love it because they don't know how it works.

it is LITERALLY impossible to look the same each time. if you turn the camera away from a scene and then back again, you will see differences, even if all conditions are identical. same camera view, no lighting changes, no dynamic objects being moved... it will never look the same.

in order for generative AI to look similar each time (note I say similar, not the same, because... repeat after me: IT'S LITERALLY IMPOSSIBLE to look the same) it will need to remember how it looked the last time you saw that exact scene. otherwise there's no guarantee it will even have the same shadowing or color temperature, especially if vital information about the lighting conditions aren't fully evident inside screen space.

You don't know shit. That is all I am going to say. just an ass talk. Because you keep comparing it to AI from chatgpt and the like. when in fact its an AI enhancer of an actual image that a artist created. it will NOT generate bullshit face every time it loads.

The devs has the ability to adjust the enhanced image to anything they want IF they want to use it . or if they want to use part of it. Nvidia said that them selves. but then comes you and the no brain people like you and spread garbage to fit an agenda.

go join the crying team. Nothing gonna change anyway.

I wish I can use it today even. That Leon face from Resident Evil is godlike. dont give a single FUCK if no one likes it.

Of course the console and AMD fanboys will cry about it. they cant do shit like that on their fav plastic box.

Every time these mfers they think are catching up to DLSS, Nvidia just stomps on them all over.

I am so glad Capcom is on board and many other studios too. for the first time since DLSS 2, I am really super excited about this. This fall can't come soon enough.

 

[GermanZepp]

YOU KNOW KOJIMA IS GONNA DROLL OVER THIS

 

[Vick]

And they manually achieved this on a Console at 60fps.
I'd obviously rather see their work progress with hardware in the future instead of seeing it replaced entirely by the same exact aesthetic we see all day long on socials, YouTube and porn websites, that we now all instantly associate with cheap trash made by nobodies in their mom's basement.

And I'll add, maybe that's why the Nvidia Path Traced version of those same cutscenes looked so bad and crude in comparison..

To have simpletons more easily blinded by their "Path Tracing looks bad in comparison".

 

[kevboard]

You don't know shit. That is all I am going to say. just an ass talk. Because you keep comparing it to AI from chatgpt and the like. when in fact its an AI enhancer of an actual image that a artist created. it will NOT generate bullshit face every time it loads.

you don't know even half a shit about what you're talking dude.

The devs has the ability to adjust the enhanced image to anything they want IF they want to use it . or if they want to use part of it. Nvidia said that them selves. but then comes you and the no brain people like you and spread garbage to fit an agenda.

you can do that with every ML image generator. input image, add a bunch of LoRAs to finetune, then generate on a massive server farm with terrabytes of memory,
and even the best ML generator, with super fine, hand tuned LoRAs will not be able to create the same "enhanced" version of an image twice.. it's literally impossible.

go join the crying team. Nothing gonna change anyway.

I wish I can use it today even. That Leon face from Resident Evil is godlike. dont give a single FUCK if no one likes it.

Of course the console and AMD fanboys will cry about it. they cant do shit like that on their fav plastic box.

I am an Nvidia user that fanboys over DLSS at every chance I get.
literally made fun of the console version of Crimson Desert just yesterday because it doesn't have DLSS Ray Reconstruction...

yet here I am telling you that you have no idea what you're talking about, and that we have clear evidence (and/or Nvidia confirmation) that this tech only sees screenspace information, and that it uses a generalised model that is not trained on specific games.

combine these 2 facts and you will (if you know how this tech works) come to the conclusion that it is nearly impossible to create a temporally stable result nor the same result twice.

 

[Cynexius]

The tech is absolutely incredible. Just because the face doesn't look exactly like the source material, it doesn't mean you should throw the baby out with the bathwater. You get real-life photorealism at relatively little cost. The technology is mind-blowing.

 

[//DEVIL//]

you don't know even half a shit about what you're talking dude.




you can do that with ever ML image generator. input image, add a bunch of LoRAs to finetune, then generate on a massive server farm with terrabytes of memory,
and even the best ML generator, with super fine, hand tuned LoRAs will it be able to create the same "enhanced" version of an image twice.. it's literally impossible.





I am an Nvidia user that fanboys over DLSS at every chance I get.
literally made fun on the console version of Crimson Desert just yesterday because it doesn't have DLSS Ray Reconstruction...

yet here I am telling you that you have no idea what you're talking about, and that we have clear evidence (and/or Nvidia confirmation) that this tech only sees screenspace information, and that it uses a generalised model that is not trained on specific games.

combine these 2 facts and you will (if you know how this tech works) come to the conclusion that it is nearly impossible to create a temporally stable result not the same result twice.

its the SAME FUCKING IMAGE. It's ENHANCED ONLY for a more photorealistic look. Jesus, you are dense. Better Hair. Better Lighting. Devs have the ability to use it however they want to enhance the image. thats a fucking Win.

Doesn't matter what you want to reply after. When the game is out this fall and support this, and the character detail stay the same, its you who will look like an idiot.

LoL @ image change every time you start the game. Hahahaha

 

[Draugoth]

The tech is absolutely incredible. Just because the face doesn't look exactly like the source material, it doesn't mean you should throw the baby out with the bathwater. You get real-life photorealism at relatively little cost. The technology is mind-blowing.

The Dragon Pattern is all wrong

 

[TheDarkPhantom]

I dedicate this one to the homosexuals

 

[kevboard]

its the SAME FUCKING IMAGE. its ENHANCED ONLY for more photorealstic look. Jesus are dense.

Doesn't matter what you want to reply after. When the game is out this fall and support this, and the character detail stay the same, its you who will look like an idiot.

LoL @ image change every time you start the game. Hahahaha

we literally saw shadows disappear because an NPC walked in front of the shadow casting mesh, and the AI losing sight of what casted the shadow, so it just thought there shouldn't be a shadow anymore now.

we saw whole color temperatures change, and shadows cast by the sun disappear because the AI has no idea where the sun is, nor how the weather conditions were.

how do you think an AI model can create a reliable result if it doesn't know the weather conditions of the game, nor which meshes on screen, and especially off acreen are sopposed to cast a shadow?

even the most fine tuned LoRAs won't fix this shit

tech illiterate.

 

[bender]

Jesus are dense.

I think Jesus was singular and since he walked on water, Id' question your density assertion.

 

[//DEVIL//]

we literally saw shadows disappear because an NPC walked in front of the shadow casting mesh, and the AI losing sight of what casted the shadow, so it just thought there shouldn't be a shadow anymore now.

we saw whole color temperatures change, and shadows cast by the sun disappear because the AI has no idea where the sun is, nor how the weather conditions were.

how do you think an AI model can create a reliable result if it doesn't know the weather conditions of the game, nor which meshes on screen, and especially off acreen are sopposed to cast a shadow?

even the most fine tuned LoRAs won't fix this shit

tech illiterate.

says the person who talks from his ass.

Yeah. I will take the DLSS5 any day even if its FHD at this image quality.

Image details look like a jump from PS3 to PS5. and its not even replacing assets. It's the same geometry.

we done here.

 

[Dirk Benedict]

From what I understand it doesn't change assets/geometry. Not a big deal.
I'm still wondering how FPS is affected...

 

[Draugoth]

two rtx 5090s used to get the result , according to DF

On two 5090s pulling 1150 watts of power or if it were to run all day, over 26kWh which is like $5 in energy per day of use.

Insane amount of power just to yassify your game.

 

[bender]

The Dragon Pattern is all wrong

And it switched it from embroidered to a print....and then the skin.

 

[Kumomeme]

lmao it kill the character model. become generic AI slop design.

 

[kevboard]

says the person who talks from his ass.

Yeah. I will take the DLSS5 any day even if its FHD at this image quality.

we done here.

you showing these screenshots as if it's an argument also shows me you don't even understand what I am saying I'm pretty sure...

because all these show is a before and after.
in order to adress what I am saying is the issue here you'd need to show a before, after, then before again and then after again, recorded at 2 separate instances and while turning the camera fully away and back again between takes.

do you know what screen space means? or what a LoRA is?

 

[crazyriddler]

Just a random thought... how will this work with nude mods? Will we be able to have hyperrealistic tits?

 

[TheAmazingSpiderMan]

It'll be hilarious when everyone's favourite console makers implements this next gen or following gen and suddenly it'll be ground breaking we all know AMD will have something like or chase something similar to this going forward.

 

[bender]

or what a LoRA is?

My favorite Tomb Raider.

 

[Nex240]

Just a random thought... how will this work with nude mods? Will we be able to have hyperrealistic tits?

Im not trying to be a gooner but It will be interesting to see if the model has any censorship in it. Not just "omg boobies" but violence/gore etc.

 

[StreetsofBeige]

Yeah. I will take the DLSS5 any day even if its FHD at this image quality.

Image details look like a jump from PS3 to PS5. and its not even replacing assets. It's the same geometry.

we done here.

Overall, I think the DLSS On pics are much better than the Off pics.

Ya, some look overdone like the Hogwarts old lady, but most pics look much better. Yes, some of the Before pics look so bad it's last gen looking.

 

[//DEVIL//]

you showing these screenshots as if it's an argument also shows me you don't even understand what I am saying I'm pretty sure...

because all these show is a before and after.
in order to adress what I am saying is the issue here you'd need to show a before, after, then before again and then after again, recorded at 2 separate instances and while turning the camera fully away and back again between takes.

do you know what screen space means? or what a LoRA is?

OMG Stahpppp. you are NOT gonna get a different image when it loads again !!!. Jesus !!

What we have seen so far is nothing but stunning and impressive. your claim has no base. bookmark this for the future when a game that support this and then lets see your claim lol.

 

[Ribi]

And they manually achieved this on a Console at 60fps.
I'd obviously rather see their work progress with hardware in the future instead of seeing it replaced entirely by the same exact aesthetic we see all day long on socials, YouTube and porn websites, that we now all instantly associate with cheap trash made by nobodies in their mom's basement.

Now show that garbage with dlss 5

 

[SlimySnake]

its the SAME FUCKING IMAGE. its ENHANCED ONLY for more photorealstic look. Jesus are dense. Better Hair. Better Lighting. Devs have the ability to use it however they want to enhance the image.

I like this tech and like you i wish i could try it out today, but devs dont have the ability to do whatever they want. Nvidia said they can change certain parameters like how photorealistic you want it to look, but the model thats being used is a singular model that is slapped on every game instead of it being trained on each game.

During the demo, the DLSS research talked through the level of granularity available. Developers don't just get an on/off switch. They get intensity controls that can be dialed anywhere, not just full strength. They get spatial masking, so they can set the water enhancement to 100%, wood to 30%, characters to 120%, all independently within the same scene. They get color grading controls for blending, contrast, saturation, and gamma. All of this runs through the existing SDK, which means studios already using DLSS and Reflex have a familiar pipeline to work with.

Basically adjust sliders on a pre-existing model which could potentially mean same assets everywhere regardless of the game.

The AI model powering DLSS 5 is a single unified model. Same model for every game. It's not trained per-title, per-face, or per-object type. It takes the raw color buffer and motion vectors as input, analyzes the scene semantics from that single frame, and enhances the lighting and material response while staying anchored to the original 3D content. It recognizes the difference between skin and metal and water and stone and foliage, and it processes each of those materials differently based on how light should interact with them.

That's not a filter. That's a fundamentally different approach to how the final image gets assembled. And it's deterministic and consistent from frame to frame, which is a hard requirement for games.

kevboard and i rarely agree on anything, but hes also right in that the way generative AI works is that you will get a different image every time. That said, nvidia says its deterministic so they are likely storing this data somewhere. We dont know enough about how the tech works though so we are all just making assumptions at this point, but at its core, generative AI has issues retaining information over longer periods of time. Ive created videos where the AI literally forgets the face after 5 seconds and adds something new to it. Im sure they have found a way around it, but its a valid concern based on what we know so far.

 

[Represent.]

Even if you don't like the faces, the difference in the environments is objectively way better with DLSS5.

Like come the fuck on what are we doing here?

 

[peish]

I trust in Nvidia. Like clockwork, Nvidia cross another tech frontier -> DLSS on/off memes -> blown away with real hands on -> Nvidia refinements -> crow eaters

This thread will be the biggest crow eating buffet in GAF history come fall 2026

I think Nvidia may even release the supers and 5090ti in conjunction with DLSS5. The memory situation seems stabilising

 

[cripterion]

Call it what you want, but

Honestly I'm not sure how I feel about this, in way it is more closer to the actual model, but I saw the other footage with close ups of Grace and while it was better, it was more like the DLSS off model with enhancements which means there's no consistency.
I need to see more of this.
Also the DLSS off is path tracing mode right?
The characters look matte and flat anyway in with PT on.
More interesting is DF is saying it ain't replacing models or textures, it's just lighting.
I mean it's possible but yet again I need to see more.

Waiting for some people in here to say the original model also look like a pornstar.

 

[SlimySnake]

This is me right now.

 

[Shogmaster]

Deep Learning Super Slop isn't where I'd thought we'd end up by 5th generation tbh.

Scratch that. Its exactly where I thought we'd end up....

And it only requires a second 5090 to accomplish? Hohoho! Sold!, Jensen you fucking human AI slop!

 

[bender]

Even if you don't like the faces, the difference in the environments is objectively way better with DLSS5.

Like come the fuck on what are we doing here?

Apparently making homeboy in blue stick out as an even more sore thumb.

 

[//DEVIL//]

I like this tech and like you i wish i could try it out today, but devs dont have the ability to do whatever they want. Nvidia said they can change certain parameters like how photorealistic you want it to look, but the model thats being used is a singular model that is slapped on every game instead of it being trained on each game.

Basically adjust sliders on a pre-existing model which could potentially mean same assets everywhere regardless of the game.

Tell that to the idiot who keeps on saying the image will look different every time you load the game.. dude has no brain...

 

[Shogmaster]

Waiting for some people in here to say the original model also look like a pornstar.

Did Capcom leave Grace to look like her? Or did they modify her face to look like Grace as her own person? Same goes with Leon face model vs Leon in game, Gideon, Sherry, Zeno, and so on.

That change took work by ARTISTS. DLSS 5 SLOPS all that work away and it takes a second 5090 to do all that Sloppening.

 

[CrustyBritches]

Do you looked at the X post above your response? What you think about that?

I looked at the high res images, what difference would that gif make?

 

[damidu]

Waiting for some people in here to say the original model also look like a pornstar.

i mean

 

[Nex240]

I trust in Nvidia. Like clockwork, Nvidia cross another tech frontier -> DLSS on/off memes -> blown away with real hands on -> Nvidia refinements -> crow eaters

This thread will be the biggest crow eating buffet in GAF history come fall 2026

I think Nvidia may even release the supers and 5090ti in conjunction with DLSS5. The memory situation seems stabilising

it's like the Sonic Cycle but for Nvidia tech.

Nvidia announce new tech > AMD/Console users downplay it > AMD/Sony announces their own version and AMD/Console users celebrate > AMD/Sony's version is released, it's called a gamechanger for finally catching up to Nvidia years later > Repeat.

This has been true for Ray Reconstruction this week too with Crimson Desert. No one will say it looks wrong or is just higher contrast once PS6 can do RR.

 

[StreetsofBeige]

Apparently making homeboy in blue stick out as an even more sore thumb.

Thats the problem with all these videos. Unless they are purposely cherry picking the bright ones, it seems every time DLSS is on it makes everything jacked up in brightness including blue man's head.

If it is supposed to improve lighting and have AI fixes to make things more realistic, why cant the lighting also be darker sometimes where pre-DLSS is too bright?

It's like the 360 gen when devs started doing all that blinding bloomy effect.

 

[Come On Tars]

Wonder how much latency this adds to the rendering pipeline.

 

[GOKUSSJ2]

 

[bender]

Thats the problem with all these videos. Unless they are purposely cherry picking the bright ones, it seems every time DLSS is on it makes everything jacked up in brightness including blue man's head.

If it is supposed to improve lighting and have AI fixes to make things more realistic, why cant the lighting also be darker sometimes where pre-DLSS is too bright?

It's a good choice that none of the NPC's are animated as Bethesda quality animations juxtaposed to a "super realistic" environment would be...something.

Every time I see that GIF, I just want to shout, "Beetlejuice, Beetlejuice, Beetlejuice".

 

[Nfactor_AF]

The cognitive dissonance from gamers and "independent" enthusiast channels on social media today has shown me how much these people are grifting their audience of followers. It's truly been something to behold. Etched forever in the halls of retardation and bad historical takes I didn't think was possible. Truly one of, if not the most embarrassing moments ever in videogame history.

 

[Represent.]

Same people who crap on DLSS5 , were praising this basic, trash ass update, Pissr 2, just hours ago.

How can you say this is amazing

And this is terrible?

We are done here. Some folks need new eyes. Or are just trolling.

 

[Ironman_Bad]

Pc civil war?

 

[bender]

Same people who crap on DLSS5 , were praising this basic, trash ass update, Pissr 2, just hours ago.

Name names.

 

[Icylobster]

The tech is absolutely incredible. Just because the face doesn't look exactly like the source material, it doesn't mean you should throw the baby out with the bathwater. You get real-life photorealism at relatively little cost. The technology is mind-blowing.

WTF happened to his face? He looks like a kid now!
WTF happened to the background?
His scars gone now too?

I get people want better graphics, but this is trading detail and craft for whatever AI decides each frame should look like. If they said this didn't have a release date, then fine, plenty of time to tune it and improve. But they sound like they want to ship it as is...

 

[cripterion]

Did Capcom leave Grace to look like her? Or did they modify her face to look like Grace as her own person? Same goes with Leon face model vs Leon in game, Gideon, Sherry, Zeno, and so on.

That change took work by ARTISTS. DLSS 5 SLOPS all that work away and it takes a second 5090 to do all that Sloppening.

The changes are made by Capcom themselves so what are you on about?

 

[Chuck Berry]

I think the newer version looks hot as fuck personally

 

[Bry0]

says the person who talks from his ass.

Yeah. I will take the DLSS5 any day even if its FHD at this image quality.

Image details look like a jump from PS3 to PS5. and its not even replacing assets. It's the same geometry.

we done here.

AI Grace be hot af tho.

 

[yogaflame]

This is good. It convince me that this is the right direction. But still you need $2000 + gpu to achieve this.

 

[Icylobster]

If you played the game you would understand Grace isn't supposed to be a Kardashian...

"But she's hot tho"

You guys are completely missing the point. Why am I even here? You can't even think beyond your groin.