Image Quality and Framebuffer Speculations for WIP/alpha/beta/E3 games *Read the first post*

[Globalisateur]

Gamersyde pics from gameplay video of Quantum Break:

http://www.gamersyde.com/gallery_25927_en.html

On this image we can find clues of horizontal sub-1080p aliasing on the board, center right between the character and the wall and also a few edges on the (low poly) vase just under the board. But those sub-1080p edges could come from lower res lighting "edges", not from framebuffer resolution, we need more clues to be more certain of anything.

Compared to the previous old pics released (that were native 1080p with obvious edges) and the recent official screenshots "bullshots" released, it's very blurry with some heavy post AA destroying all details + DOF + strong grainy filter and probably sub-1080p just judging from the obvious lack of clarity.

The official screenshots look indeed different, much better than those gameplay pics, IMO it's using the same gameplay engine but only downsampled from higher res like 4K.

 

[Laa-Yosh]

film grain can definitely be regarded as a noise source. It's a highly randomized aberration to the source data.

Now you're being way too technical... remember, I'm about half an artist

 

[COPS N RAPPERS]

I've seen the capture from gamersyde, i got nothing from it. the youtube version looks better and even still. https://www.youtube.com/watch?v=c-73i1k47_Q#t=108

Aliasing on distant objects is what i would check for, i saw subtle movement but that's it.

 

[Rangers]

Err, that Gamersyde video is actually filmed with a camcorder off screen. That seems like a horrible way to analyze footage.

Not aware of anything QB higher quality than crap youtube though. Nothing at Gamersyde besides the offscreen.

Does look very "soft"

 

[taisui]

Err, that Gamersyde video is actually filmed with a camcorder off screen. That seems like a horrible way to analyze footage.

Not aware of anything QB higher quality than crap youtube though. Nothing at Gamersyde besides the offscreen.

Does look very "soft"

Pretty sure that adds a lot of post processing and motion blur

 

[AlNom]

Oops.

----

The temporal ghosting is quite tempura.

----

Gotta fix some of the Z-fighting decals on the ground and such.

 

[Starx]

Captured from 1080 direct feed (youtube).

 

[Starx]

Very hard to tell but i think native resolution is 1080

 

[HTupolev]

Very hard to tell but i think native resolution is 1080

Really? Based on the area you boxed, I'm seeing something vaguely in the ballpark of 8/11, though my (admittedly very dubious) count on the lower edge was 7/11. We're possibly looking at <800p.

 

[AlNom]

Never examine specular res.

 

[COPS N RAPPERS]

Can't tell anything from the videos. whatever resolution it is, It definitely seems far away from 720p. distant objects that are normally sensitive to resolution like power lines, cables, and foliage, sustain enough shape in the footage.

 

[Rockster]

Has anyone done a comparison between 1600 x 900 and say 1332ish x 1080? I remember reading somewhere that people were more sensitive to vertical res than horizontal (but not sure that's true). Obviously results would vary by content but anyone know how these compare in general?

 

[HTupolev]

Never examine specular res.

I'm curious, what exactly breaks down?

Is texel density and texture filtering on normal maps (and similar such stuff) the big issue?

 

[LightHeaven]

Gamersyde pics from gameplay video of Quantum Break:

The Gamersyde video, albeit the high bitrate is offscreen footage, kinda unfair to compare with direct feed screenshots and demand the same sharpness.

That low poly vase that you spotted, I feel like it's tessellated, maybe it's from the video or that distortion effects, but I have the impression it changes its geometry when it's about to leave the camera view. When I was watching the conference I think I saw some bugs too, where some objects appeared to toggle their geometry quality.

 

[zed]

I'm curious, what exactly breaks down?

Is texel density and texture filtering on normal maps (and similar such stuff) the big issue?

nothing breaks down
it just goes from>10000000 to 0.002 within a pixel thus its not really valid for measurement cause on A->B it can be any number from 0.002 -> 10000000 i.e. not stable (unless the cam and light cource is completely static) so it breaks down cause the the range has too many variables that change on minute differences

 

[Shifty Geezer]

Now you're being way too technical... remember, I'm about half an artist

Think of it like audio noise. Tune in to static on the radio and you get white noise. Now run it through an equaliser. You can have low, rumbly noise, or high, watery noise, very different in sound but they're still noise. Where the data points are derived from randomness and don't form any pattern of values, it's noise. One can point to Perlin Noise too. 100% algorithmic, but it's still noise. Specific film grains are just different flavours of noise. Also, digital camera shouldn't have such noise qualities. It should be pure RGB noise per pixel, and that would be the type of noise added to games to simulate a digital camera.

 

[ultragpu]

Silent Hill P.T.

Looks native with some post AA anyone?

Halo 5 Beta

http://www.gamersyde.com/news_gc_halo_5_multiplayer_beta_screens-15665_en.html

 

[HTupolev]

Pretty certain the Halo images are native, but they're using Post-aaaaaaaaaaaaaaaaaaaaaah!

There's a missed edge on the right of the staircase towards the back left of the first image there. They need to work on the specular aliasing too...

Where? Even in the first image, I'm not seeing any 1/1 stairstepping.

The second image is loaded with ~2/1 stairsteps, though. Looks like badly-aliased 4K.

edit: -.-

 

[Globalisateur]

Where? Even in the first image, I'm not seeing any 1/1 stairstepping.

The second image is loaded with ~2/1 stairsteps, though. Looks like badly-aliased 4K.

edit: -.-

Those last 2 Halo 5 pics are indeed native 4K images downsampled to 1080p. And yes even the first image has plenty 2/1 stairsteps...

This one posted 2 pages ago is a native 4K image downsampled to 1440p:

Their postAA seems to miss some very nocticeable highly contrasted edges, weird...Are they even using any AA?

 

[Laa-Yosh]

Also, digital camera shouldn't have such noise qualities. It should be pure RGB noise per pixel, and that would be the type of noise added to games to simulate a digital camera.

And yet, there are software tools on the market which offer look-up tables and other tools to turn your digital camera footage into an almost perfect recreation of the most commonly used film stocks. There's a demand for such tools and it's obviously driven by artistic goals.

Sometimes pure technology is just not the answer.