Real-Time radiosity

Real-Time radiosity

Re: Real-Time radiosity Posted by reaper47 on Thu Mar 1st 2007 at 1:06pm
reaper47
2827 posts
Posted 2007-03-01 1:06pm
reaper47
member
2827 posts 1921 snarkmarks Registered: Feb 16th 2005 Location: Austria
Maybe you remember that the main reason Valve didn't use real-time shadows a la Doom3 was them being a) too sharp b) too dark (no light bouncing/radiosity)

Now, some of you might have noticed how obsessed I am with lighting. Needless to say, I was quite excited when I found this OpenGL demo featuring real-time radiosity that runs on my Radeon 9800 Pro! Sure, this is rather low-quality and resolution (at least in the mode I can run it on) but think of this in the hands of a professional studio and with better hardware support. I think that Crysis will feature a similar technology and logically, Valve should be able to integrate this into Source within the coming year(s). This is one of the last boundaries besides ray-tracing and a few physics-based effects.

User posted image

Awesome!

If you know any other demos of that sort, please point me to them!
Why snark works.
Re: Real-Time radiosity Posted by Naklajat on Thu Mar 1st 2007 at 2:37pm
Naklajat
1137 posts
Posted 2007-03-01 2:37pm
Naklajat
member
1137 posts 384 snarkmarks Registered: Nov 15th 2004 Occupation: Baron Location: Austin, Texas
Geomerics' Enlighten realtime radiosity was recently added to the UE3 integrated partners program (meaning dev studios can license the tech, but it won't be normally included with UE3 licenses). I agree that it's an exciting technology for the future of games, especially given how amazing some of the new games with just dynamic shadows and static (or even faked) radiosity already look .

o

Re: Real-Time radiosity Posted by reaper47 on Thu Mar 1st 2007 at 2:54pm
reaper47
2827 posts
Posted 2007-03-01 2:54pm
reaper47
member
2827 posts 1921 snarkmarks Registered: Feb 16th 2005 Location: Austria
Re: Real-Time radiosity Posted by Crono on Fri Mar 9th 2007 at 3:32am
Crono
6628 posts
Posted 2007-03-09 3:32am
Crono
super admin
6628 posts 700 snarkmarks Registered: Dec 19th 2003 Location: Oregon, USA
Yeah, but notice the poor sampling around the edges of the shadows?
That's why I'd suggest a heuristic like shadow mapping or shadow volumes (or both) with some blurring based on light properties.

A common issue with global illumination techniques is that scratchy dotted look of sampled areas.

Also, if you notice in that scene, something that is missing is specular effects and caustics. Because radiosity methods do not account for these "special cases" in those instances you have to actually cast some rays and see where they go. That's why you commonly see ray tracing used in conjunction with this.
Blame it on Microsoft, God does.
Re: Real-Time radiosity Posted by reaper47 on Fri Mar 9th 2007 at 11:04am
reaper47
2827 posts
Posted 2007-03-09 11:04am
reaper47
member
2827 posts 1921 snarkmarks Registered: Feb 16th 2005 Location: Austria
It's still a huge leap from stencil shadows a la Doom 3. Man, if this was introduced to HL3 (or Episode 3!)...
Why snark works.
Re: Real-Time radiosity Posted by Crono on Fri Mar 9th 2007 at 7:46pm
Crono
6628 posts
Posted 2007-03-09 7:46pm
Crono
super admin
6628 posts 700 snarkmarks Registered: Dec 19th 2003 Location: Oregon, USA
No it isn't. Stencil Shadow Volumes is a very advanced technique and can be used in combination with a global illumination technique to produce the shadows.

You have to understand that you don't want to calculate everything all the time, you just want it to look really nice.

Calculating all the light all the time, while it will run in a small situation on modern cards will not fly in large open areas. There are things you need to do to limit that. Of course, something you can do, is in your lighting algorithm, say when you hit some point and you need to cast shadows you can begin doing stencil shadow volumes.

I think people are really annoyed by those because of Doom 3, but they should realize, that D3 shadows had no aliasing ... self shadowed everything and were dynamic. Just because they don't like the hard edge look they damn the entire process which is ridiculous. You can blur the edges based on the intensity and distance the object is from the light. Make a hybrid algorithm with some of the soft shadow techniques.

To really do this stuff correctly, making algorithms work together to make something as efficient as possible is the goal (or at least one of them). Just sitting down and going, "Wow, I wish games did that" is fine, but there are definitely implications! :smile:
Blame it on Microsoft, God does.
Re: Real-Time radiosity Posted by reaper47 on Fri Mar 9th 2007 at 8:30pm
reaper47
2827 posts
Posted 2007-03-09 8:30pm
reaper47
member
2827 posts 1921 snarkmarks Registered: Feb 16th 2005 Location: Austria
Well, this thread is on the level of a group of regulars chatting about what they read on the interwebs. I do wish games did that today, that's all I can do as a consumer (vs being a devloper). I find these insights very interesting, though.

You seem to be a bit upset with people complaining about D3's edgy shadows and while I understand that from the programmer's point of view I think you have to consider the history of this discussion. It was mainly HL2 (the Source engine) vs the rest of the games. Doom3 had been advertising its shadow technology for years and Valve had to explain to the crowd of highly-priced next gen graphics cards owners that they won't be using a core technology of the time.

It could have easily gone into the other direction (and actually has) that people accuse HL2 of having "bad lighting" because light isn't done in real time for the world. While many others thought that HL2 had a very good reason for not following this trend because with the old lighting methods real-world architecture looks better than D3's. I feel like Valve's decision is much more vulnerable than Doom 3's. That's why I tend to defend them. I could imagine from a programmer's point of view reading things like "Doom 3's lighting sucks because of the hard edges" must be frustrating because it's ignorant.

I love D3's lighting and HL2's. I just think, from an artistical standpoint, the technologically outdated method of HL2 gives better results for the environments you find in the game. That's my point of view on this. I think D3 got all the praise in the world and well-deserved.

Global illumination breaks those barriers completely which is why I think it's so exciting. I do hate running vrad for an hour only to realize that the lightmaps are bugged around a spotlight. Oh yes, I do. :smile:
Why snark works.
Re: Real-Time radiosity Posted by Crono on Sat Mar 10th 2007 at 5:21am
Crono
6628 posts
Posted 2007-03-10 5:21am
Crono
super admin
6628 posts 700 snarkmarks Registered: Dec 19th 2003 Location: Oregon, USA
I'm not upset, I just find it odd that, in general, if someone who isn't familiar with the technique gets introduced to it in some specific manner think that's the only way it could be used. Which couldn't be further from the truth. I've just been using D3 as an example because in these specific discussions people are bringing it up as a bitching point.

I don't think we're quite ready for full-on GI solutions. It just doesn't realistically run in real-time. That's why different lighting algorithms are needed based on distance away. That's something that a lot of developers seem unwilling to do.
Blame it on Microsoft, God does.