"Shadows aren't really that important"

[Saem]

I think in this next gen, engines will become less disposable.

Shadows add content for free. Well, "free", you basically get a lot of effect in a persistent fashion. Now start make them dynamic, global, add self shadow, soft shadows... all of a sudden artists spend less work compensating for something the game/console can't do. And all this on a few programmer's pay, who produce code which will likely be reusable, instead of artists pumping out models and textures, which will likely be less so, per game.

Shadows like all procedural content, is a win. I don't see how, budgets, financially speaking, act against you in this matter.

 

[KimB]

On a console you are sitting far from the TV screen - Will the majority be able to tell the difference between hard and soft shadow types? (ignoring the self shadow issue).

Nextgen consoles will be widely used on nextgen tv's (HDTV).

I don't see how that affects hard shadows vs. soft shadows.

 

[Leto]

On a console you are sitting far from the TV screen - Will the majority be able to tell the difference between hard and soft shadow types? (ignoring the self shadow issue).

Nextgen consoles will be widely used on nextgen tv's (HDTV).

I don't see how that affects hard shadows vs. soft shadows.

Big highres HDTV's a lil distance away would reveal as much detail as a nearer pc monitor.

 

[KimB]

Big highres HDTV's a lil distance away would reveal as much detail as a nearer pc monitor.

Who said big HDTV's are going to be popular? They're expensive as hell, and will continue to be for some time to come.

 

[karlotta]

IMO the budget issue is product to market. With Xbox comming fast, your time is running out for a game at launch. MS wants lots of AAA games at launch.

 

[KimB]

IMO the budget issue is product to market. With Xbox comming fast, your time is running out for a game at launch. MS wants lots of AAA games at launch.

I don't see shadows as really adding any extra development time, though. The budget that they affect is the performance budget. They're easy to implement, they just take quite a bit of processing power. As long as they're in the engine from the start, and therefore affect the triangle count and overdraw budgets of the game as development is proceeding, they should add no significant extra development time or, therefore, cost.

 

[DemoCoder]

I can definately tell the difference between hard and soft shadows, just like you can tell if a scene in a movie or photograph is lit too harshy or if you monitor lacks contrast ratio and crushes blacks.

Human beings are very good at picking up imperfections. You might not be able to point out "aha! there's a hard shadow", but you will look at the picture and feel "something isn't right."

I don't think implementing shadows are any harder than what developers are doing, like Valve, authoring thousands of material shaders. Shadows eat performance, I don't think they are significantly more time consuming than implementing other parts of a game's rendering.

 

[KimB]

Well, I think it's obvious that soft shadows are the way to go in the future, but until we get robust shadowing techniques (ex. shadow mapping) that allow for high-performance soft shadows, or performance increases enough to make them viable, hard shadows are definitely a great first step.

Of course, in the very long term, we don't just want "soft" shadows, but a more complex rendering method where one also considers diffuse emmission from surfaces (soft light).

 

[Leto]

Big highres HDTV's a lil distance away would reveal as much detail as a nearer pc monitor.

Who said big HDTV's are going to be popular? They're expensive as hell, and will continue to be for some time to come.

Oh for christ sake, I'm not saying he hasn't got a point, I'm just adding a lil thought to consider.

 

[OICAspork]

Shadows are only going to get more important in consoles.

I dug out a link to this thread in the console forum with pictures of Resident Evil 4 for gamecube. There is an animated gif that demonstrates some sort of soft self shadowing. Watching the actual trailer with the torches flickering is pretty spectacular...

In games that have a realistic art-direction games, accurate shadowing is going to become increasingly important, especially next gen.

Edit... here is a link to the most recent trailer.

 

[Reverend]

The developer I spoke with does not have "special deals" with any of the console manufacturers. His base target is the PS2. His "higher ups" priority is the console, not the PC.

I have personally never regarded shadowing as "important" in every console game I played. It has to said I haven't played that many console games however.

Perhaps this developers' thoughts mirrors mine in one aspect -- there appears to be more focus on gameplay, rather than graphics, in console games. While I wouldn't say this (gameplay > graphics) doesn't hold true wrt PC games, I get the distinct feeling that the focus on graphics is higher/more important for PC games (this is related to what I said earlier about the "ability" of the PC to progress faster than the consoles). As I read more and more console game reviews and PC game reviews, it is evident to me that is quite true -- the focus on graphics is higher in PC game reviews compared to console game reviews. Some kinda "mindset" for (PC and/or console) game reviewers perhaps?

B3D being what it is (mainly PC), I can understand the comments given wrt the importance of shadowing. I'm just providing the thoughts of a game developer that has to make a game that appeals to both platforms. He starts out on his game engine with this being a priority. That is perhaps an important consideration that may not have been considered by some here. We're not talking id Software

 

[AlNom]

The developer I spoke with does not have "special deals" with any of the console manufacturers. His base target is the PS2. His "higher ups" priority is the console, not the PC.

There goes my Itagaki theory. Stick shadows for characters in his games are too obvious.

 

[noko]

Creating a realistic as possible enviroment for certain type games I think requires good shadowing while others type games it wouldn't make much of a difference. For example; a surfing game would probably benefit very little with shadows while an indoor game like DoomIII is utterly transformed by it.

Getting involved in the game or game play I think has many different aspects. A chess game involvement is strategy in which graphics has virtually no real impact. DoomIII is a totally different situation where the graphics takes you into the game. Which one is more involved with game play? Both are, different type elements are needed for different types of games. For some games Shadows make a huge difference while other type games it is unimportant.

 

[991060]

If the developer's stance can be interpreted as "If you can't make sth good, don't make it in the first place", I'll agree with him. Take a look at the shadows in 3DMark05, they're not good shadow in every aspect.

 

[Dave B(TotalVR)]

When i first played the leaked Alpha of Doom 3 my machine couldn't really handle the graphics so i started to experiment with the settings. I ended up disabling stencil shadows which had a rather minor visual impact compared to disabling normal mapping or specular highlights.

Which in turn ruins some very cool scenes in the game. One of my favorite sequences was during a section of the game where there were no opponents to fight (which I thought was quite cool, as you spend quite some time wondering just when they'll come....), and an imp crawled across a skylight. I saw the shadow on the ground first, and almost shot at it.

Global shadows really do add a measure of depth to the game, and I think it's something that we should be able to demand of nearly all new games in the next couple of years. Beyond that, we should expect soft shadows and such. Yes, they're computationally expensive, but they add so much...

Im totally with Chalnoth on this one but I'd like to add that there should be more research and investigation into stencil shadows or shadow volumes in general - which are the most computationally correct form. Their problem, of course, is generating depth dependant soft edges like we all know real shadows posess. I just know there is a way around it, ive got fragments of ideas in my head and Ive been meaning to sit down sometime and work it all out, but I can tell its gonna give me a big headache.

 

[KimB]

Im totally with Chalnoth on this one but I'd like to add that there should be more research and investigation into stencil shadows or shadow volumes in general - which are the most computationally correct form.

Personally, I'm much more for the implementation of an irregular z-buffer for shadow mapping. Do a search on these forums if you don't know what I'm talking about.

 

[Frank]

Im totally with Chalnoth on this one but I'd like to add that there should be more research and investigation into stencil shadows or shadow volumes in general - which are the most computationally correct form. Their problem, of course, is generating depth dependant soft edges like we all know real shadows posess. I just know there is a way around it, ive got fragments of ideas in my head and Ive been meaning to sit down sometime and work it all out, but I can tell its gonna give me a big headache.

Why not project 2 cones: one at the outer limit of the penumbra and one at the inner? Sure, it would be even more geometry, but the tricky work (finding the edges and such) only has to be done once. Although I have no idea how you could calculate the gradient that way, so it is probably a stupid idea.

 

[KimB]

Why not project 2 cones: one at the outer limit of the penumbra and one at the inner? Sure, it would be even more geometry, but the tricky work (finding the edges and such) only has to be done once. Although I have no idea how you could calculate the gradient that way, so it is probably a stupid idea.

Well, I think the problem with this approach is that the first cone is easy to compute. But for the second cone you're going to have to do something that's object-dependent, such as move the light source towards or away from the object. This may make it a bit more challenging to do shadow volume extrusion on the GPU. As for the gradient, yeah, I have no idea...but this is going to be hard for any screenspace technique, and any light-space technique will suffer from aliasing, so there's no perfect solution.

 

[davepermen]

shadow volumes have no future due 2 reasons:

scalability (small object at top throwing small shadow on floor, but you fill the whole screen from top to bottom at least twice (if it's a grid the overdraw gets hell of a lot more) / performance estimation. the rendering speed greatly depends on where you are, and where the shadow-volumes go to. this is not directly linked with the scene, and can lead to huge fps drops.. ugly.

robustness. yes, they are a robust solution to flat shaded triangular meshes that don't have any transparency. if your objects can have smooth shading, or even subsurface scattering effects, shadowvolumes create ugly edges. and if your object has transparencies, even just a colorkeyed texture, or some alpha map, it completely fails.

for colorkeyed shadows, independent on geometry complexity, with scalability, shadowmaps work great.

you don't want to have a tree with each leafe polygonal, extruding a full shadow volume down to the floor.... you really don't..



the only correct solution for translucent objects, sharp shadows (does not mean hard shadows, but precious ones, even correct soft ones), no mather what geometry, and with good scalability and cost estimation, you'd need raytracing..

but to get there, thats a fully different topic

 

[KimB]

shadow volumes have no future due 2 reasons:

That's just fine, but shadow maps have similar rendering characteristics to shadow volumes. Granted, they do change the way the shadows are applied, but I don't know if that's terribly significant to performance scaling.

the only correct solution for translucent objects, sharp shadows (does not mean hard shadows, but precious ones, even correct soft ones), no mather what geometry, and with good scalability and cost estimation, you'd need raytracing..

Nah, an irregular z-buffer with shadow mapping (basically shadow mapping in screen space) should be good for this task, if hardware developers decide to support it.