Next Step in 3D Graphics

[EnderAndrew]

The commercial for NCAA 2006 is playing on my TV and it keeps cutting from actual video footage of a college football game and the video game. In these cuts one thing keeps jumping out at me. It seems the greatest disparity between high-res graphics right now and photorealism is focus.

 

We only seem to be paying attention to texture details and lighting.

 

Why is it that we haven't developed a toolset, or that DirectX hasn't focused on a means to focus the camera on a primary object, while background objects lose focus?

[Loof]

But is that a feature we realy want? At least for gameplay I can imagine it being very irritating for the player if the player is focusing on one object but the engine desides that he is focusing on something else...

So far I don't think anyone has made a telepathic game engine

 

Although I admit that it would definetly be a cool feature to have for cutscenes and such. and it might be usefull as a targeting indictor (by unfocusing all entitys that are not being targeted).

[EnderAndrew]

Anything reasonably close to the main character in third person view would still be in focus. In first person view, focus would be more difficullt to determine.

[metadigital]

But is that a feature we realy want? At least for gameplay I can imagine it being very irritating for the player if the player is focusing on one object but the engine desides that he is focusing on something else...

So far I don't think anyone has made a telepathic game engine

 

Although I admit that it would definetly be a cool feature to have for cutscenes and such. and it might be usefull as a targeting indictor (by unfocusing all entitys that are not being targeted).

<{POST_SNAPBACK}>

Use the same functions that are in use now, for choosing which player kicks the ball / tackles / etc.

[EnderAndrew]

Sounds good to me.

[alanschu]

I don't think it would work as well in a sports game, except perhaps highlights and whatnot.

 

 

For a more cinimatic game, it would probably be good. Although I think it might be restricted more towards cutscenes.

[EnderAndrew]

I'm not talking about drastic changes in focus.

 

Here's a real photo.

Here is a screenshot.

In the screenshot, there are a few things that keep us from photorealism.

 

With next-gen consoles and PC gaming we are going to be able to get full anti-aliasing, lose the jaggies, and get higher res textures, and better models.

 

When you improve all of those, the only thing really left is focus.

[alanschu]

I still think we need some more polygons.

 

Polygons has always been the focus for me. I can look past jaggies and whatnot (even at 1024x768 with no FSAA, I only notice them if I'm looking for them). But when someone has a blocky face or something, then ugh.

[EnderAndrew]

Again, next-gen consoles will feature much better models. Look at the Unreal 3 tech demos, or screenshots from any of the next-gen games we've seen. I accept that as a given.

[alanschu]

I'm not saying that they don't have really nice models.

 

But I still think we could use some more polygons

[EnderAndrew]

From what I've been reading, using normal bump-mapping and newer techniques, you can make much better looking models with fewer polygons.

[alanschu]

Actually, the per-pixel effects can do an excellent job.

 

In fact, Half-Life 2 is a stellar example of this. Compared to Doom 3, it's polygon count is actually quite a bit lower. However, with their mad use of bumpmapping, I find the game actually looks better (aside from the shadows). I remember seeing some side by side comparisons of Half-Life 2's graphics with textureless polygons, and textured polygons, and you'd have no idea the polygon count was that low.

 

So I guess I stand corrected...to an extent.

 

I wonder if it's a good or bad thing to have the perpixel effects, combined with increased polygons?

[tmp]

I'm not talking about drastic changes in focus.

 

Here's a real photo.

Here is a screenshot.

In the screenshot, there are a few things that keep us from photorealism.

 

With next-gen consoles and PC gaming we are going to be able to get full anti-aliasing, lose the jaggies, and get higher res textures, and better models.

 

When you improve all of those, the only thing really left is focus.

<{POST_SNAPBACK}>

If you tried to pass the top picture as game screenshot, the first comment you're going to hear is "omg how could they use such hideously low res textures for the background" ... because that's what the focus boils down to, blurry mess

 

Furthermore, games tend to simulate human vision which handles keeping things in focus much better than camera. What you see in reality is actually closer to the bottom picture than to the one on top...

[metadigital]

Don't forget dynamic range lighting, as in Half-Life 2: The Lost Coast. I hear there are other developers working on similar improvements, so I would expect Quake V to have some sort of variable bloom and darkness effects over distances.

[alanschu]

Looking forward to Lost Coast

[Azarkon]

The commercial for NCAA 2006 is playing on my TV and it keeps cutting from actual video footage of a college football game and the video game.  In these cuts one thing keeps jumping out at me.  It seems the greatest disparity between high-res graphics right now and photorealism is focus.

 

We only seem to be paying attention to texture details and lighting.

 

Why is it that we haven't developed a toolset, or that DirectX hasn't focused on a means to focus the camera on a primary object, while background objects lose focus?

<{POST_SNAPBACK}>

 

Realistic and accurate depth of field approaches do exist:

http://features.cgsociety.org/story.php?story_id=2917

 

However, they're not fast enough for games. There have been recent developments in this area, however, with the possibility of a major breakthrough:

 

http://www.ati.com/developer/shaderx/Shade...dSimulation.pdf

 

http://csdl2.computer.org/persagen/DLAbsTo....1109/38.486676

 

True depth of field in games is not far off, and it is definitely in development. One problem with using DoF though is that what the eye sees as DoF in real life might not match what the computer generates as DoF in a virtual environment. This could lead to the player to notice the fake DoF and be less immersed than if DoF did not exist. This isn't so much a problem in cinematography as it is in interactive worlds because cinematography always allows you to adjust the shot until it's just right and what you want, and as with photos, it's not meant to be first person. But imagine your eyes focusing on the screen in a first person game: true DoF isn't just a simulated blurring in reality, and your eyes CAN adjust its focus, so if I put a coke bottle in front of you with DoF blurring the background, it'll be as if your eyes were focusing on the bottle. But now try adjusting your eyes to focus on the background - that isn't possible on a computer but is in rea life, and may therefore result in a less realistic environment.

 

So in conclusion, we're getting there, but there are giant hurdles to overcome.

[EnderAndrew]

Thanks. That's really good info.