DLSS 5 will be the worst thing to EVER happen to video games

Retros_x said:

The right also throws out the artistic intent and design. For pure realistic games this might be cool, but I would be pissed if I would be an art designer or character designer for a video game company.

Click to expand...

We don't yet know how customisable this will be when implemented by game companies into their own games. There must be some way to enforce consistency, otherwise the same character would wind up looking like different people from scene to scene, and such tools may include directions as to what direction the system takes when changing the visuals.

I really can't tell the difference except that the one on the right looks slightly better lit.

briggart said:

To me, the one on the left looks very flat, the one on the right looks it was lit by somebody who understands how to properly use light to shape a subject (and maybe with some makeup applied to the model).

Click to expand...

It looks like someone who understand portrait photography and don't have a single clue about cinematography or natural light.



Let's begin with the A circles because they're kinda crucial. Itr's a foggy, dimly lit scene as one would expect in Resident Evil. The DLSS removes that completely, making everything in focus and removing a lot of atmosphere markers.

B. Soft light as one would expect down in the sidestreet. But no, DLSS relights here with a ring light in her face and with brighter refelections indicating that what lights here is closer and to the frond and a little up. It's good for photography, which this isn't. Also her hair has highlight from all angles like it's a bit of a backlight. So she's lit from three different sources at equal intensity and colour.

C. The yellow light from the window is reflected on the boxes making it glow a little. Here DLSS removes light instead of adding it. The light from inside is somehow blocked by the glass pane.

D. I have no idea how DLSS decided that a car under a roof in the fog should have a a bright sidelight hitting the front of it.


For another example, the workshop scene in Starfield. There are no structural shadows left, everything is bright and visible. Yet again, the light is not coming from any source you can see and all is coming in the direction from the camera.




Someone also pointed out that the before images all look blander and less detailed than it does on their machines, so it's not suing all of the rendering available probably because it's less confusing for DLSS to apply it's algoritms if the source is plainer.

Paul Farquhar said:

Err, yes? There might be all sorts of ethical and practical issues around this software, but all I see are two CGI images. The left hand image is rather dark, the right hand image is better lit.

Click to expand...

The left image has actual artistic direction. The right image splatters random and arbitrary brightening all over it.

Also it's a Resident Evil game. It's supposed to be dark.

Gladius Legis said:

The left image has actual artistic direction. The right image splatters random and arbitrary brightening all over it.

Click to expand...

I can’t tell that by looking.

Gladius Legis said:

Also it's a Resident Evil game. It's supposed to be dark.

Click to expand...

You didn’t provide any context. There was nothing to indicate the image was supposed to be dark or horror themed. And the left image is a pretty bad “dark” image - you realise human colour vision doesn’t work in low light?!

She also has sausage fingers. I would guess the left hand image was deliberately chosen because it was bad, in order to make someone’s product look good, but there isn’t any evidence for that in the images themselves.

I certainly can’t see anything in the images that is directly threatening. I would rather play an isometric game myself. I wouldn’t buy either.

What exactly is your problem here? Someone is selling something you don’t think is very good?

Paul Farquhar said:

I can’t tell that by looking.

Click to expand...

Well, I can.

Paul Farquhar said:

You didn’t provide any context. There was nothing to indicate the image was supposed to be dark or horror themed. And the left image is a pretty bad “dark” image - you realise human colour vision doesn’t work in low light?!

Click to expand...

You realize there are different degrees of "dark", right?

nyvinter said:

It looks like someone who understand portrait photography and don't have a single clue about cinematography or natural light.

View attachment 432280

Let's begin with the A circles because they're kinda crucial. Itr's a foggy, dimly lit scene as one would expect in Resident Evil. The DLSS removes that completely, making everything in focus and removing a lot of atmosphere markers.

B. Soft light as one would expect down in the sidestreet. But no, DLSS relights here with a ring light in her face and with brighter refelections indicating that what lights here is closer and to the frond and a little up. It's good for photography, which this isn't.

C. The yellow light from the window is reflected on the boxes making it glow a little. Here DLSS removes light instead of adding it. The light from inside somehow sis blocked by the glass pane.

D. I have no idea how DLSS decided that a car under a roof in the fog should have a a bright sidelight hitting the front of it.


For another example, the workshop scene in Starfield. There are no structural shadows left, everything is bright and visible. Yet again, the light is not coming from any source you can see and all is coming in the direction from the camera.


View attachment 432282

Someone also pointed out that the before images all look blander and less detailed than it does on their machines, so it's not suing all of the rendering available probably because it's less confusing for DLSS to apply it's algoritms if the source is plainer.

Click to expand...

Well, this provides more context than the image in the OP. Some choices are indeed puzzling.

Aeson said:

Recency bias? I get COVID was pretty recent compared to the other things, but to a teen, that's a hundred years ago. What was a hundred years ago might as well be hundreds of years ago. AI is now. AI is the future. I wonder if the poll asked about different events if the answers would be the same.

If you could prevent one of these things?
Invention of generative AI
Mass deportation
COVID-27
WW III

Click to expand...

It would be hard to vote as we do not know what will be the impact of the 1st choice and rest are still to happen(hopefully not) and in what aspect will the rest happen?

nyvinter said:

It looks like someone who understand portrait photography and don't have a single clue about cinematography or natural light.

View attachment 432280

Let's begin with the A circles because they're kinda crucial. Itr's a foggy, dimly lit scene as one would expect in Resident Evil. The DLSS removes that completely, making everything in focus and removing a lot of atmosphere markers.

B. Soft light as one would expect down in the sidestreet. But no, DLSS relights here with a ring light in her face and with brighter refelections indicating that what lights here is closer and to the frond and a little up. It's good for photography, which this isn't.

C. The yellow light from the window is reflected on the boxes making it glow a little. Here DLSS removes light instead of adding it. The light from inside somehow sis blocked by the glass pane.

D. I have no idea how DLSS decided that a car under a roof in the fog should have a a bright sidelight hitting the front of it.


For another example, the workshop scene in Starfield. There are no structural shadows left, everything is bright and visible. Yet again, the light is not coming from any source you can see and all is coming in the direction from the camera.


View attachment 432282

Someone also pointed out that the before images all look blander and less detailed than it does on their machines, so it's not suing all of the rendering available probably because it's less confusing for DLSS to apply it's algoritms if the source is plainer.

Click to expand...

For the record, this is a well-done post that communicates the drawbacks of the algorithm clearly. I appreciate you taking the time to do this.

Other posters could do well to reflect on your method of conveying criticisms clearly and with a minimum of unnecessary rage.

nyvinter said:

For another example, the workshop scene in Starfield. There are no structural shadows left, everything is bright and visible. Yet again, the light is not coming from any source you can see and all is coming in the direction from the camera.


View attachment 432282

Someone also pointed out that the before images all look blander and less detailed than it does on their machines, so it's not suing all of the rendering available probably because it's less confusing for DLSS to apply it's algoritms if the source is plainer.

Click to expand...

you know the DLSS 5 on makes the NPCs actually look better imo