Someone showed me this recently and its really impressive. It's exciting to think about what sort of lighting tech we'll be using in the near future for games.
Really nice stuff. We're using tech now that lets you have emissive surfaces actually be a source of light for the rest of the level and that sort of stuff can get pretty fun when creating realistic lighting.
Well I suspect they're doing something like updating the intensity/color of light probes scattered around the level, but the workflow vid makes it pretty clear that it isn't just specific areas acting as direct emitters, instead they seem to be actually emulating bounce light.
I'm sure someone is going to jump all over this if they haven't already. Considering it runs inside Unreal3 and on PS3/360's I don't see why anyone would pass on this, unless the price is really high, which could be =/
Well, just because it 'runs' it doesnt mean it runs well, the only thing they have showed so far is very small and enclosed areas. If it ran decently in larger areas with more stuff they would most probably have showed that. Im not saying its not awesome, it is, I just dont think that the hardware is ready for it on any grand scale, but for lets say Alone in the Dark or something, it could probably be used.
:poly115:
Really impressive stuff. Like hellobuddy mentioned, I would be interested to see how this handles in large open areas. But the tech is awesome.
Consider looking into what's known as "deferred shading", it's not exactly new, but in the last few years alot of high end engines are switching the lighting pipeline over to the deferred shading model. It allows for many lights with MUCH less impact.
Getting back to this, when I first saw it, my immediate thoughts were along the lines of "I want someone to make a mystery survival horror with this...that doesn't suck/isn't gimmicky"...
The outside area, where the sun casts shadows...really got my attention. It would make camping in shadows in tactical fps games really interesting. You wouldn't have folks camping in the same spots because they would need to move around to different areas depending on the sun's place in the sky. Games such as thief could REALLY receive a boost.
yeah that looks ace, but i have my suspicions about how much of what is shown in video is possible to run within a complex game engine on 360. even so would be ace to make beautifull simple puzzzle game or something similar even if a full blown fps wouldnt
Sweet, lighting, but am i the only one who noticed the lack of normal and specular?
Now show me a demo of a modern building interior with lots of shiny bumpy things using that lighting. If they can do that then they will have allot more sales. As is it looks like a dynamically lit Prince of Persia game.
deferred lighting is definitely the way of the future . 100s of dynamic point based lights with no performance loss, just place them where you the sunlight hits a surface and voila, realtime radiosity. I guess with this system the brightness of the points is determined dynamically by the sunlight projection around it?
Anyways the future is bright for artists, no more baking lightmaps and creating lightmap uvs . Deferred renderers also work great with SSAO.
Howmany games are out today that support deferred rendering? I only know of Crysis and GRAW&GRAW2 for pc. I think stalker has one too but i'm not sure.
Thegodzero: normalmaps arnt a problem with a deferred renderer, they actually look alot better becuase alot more specular points at thesame time are possible.
Deferred shading sounds nice. Apparently has problems with transparent surfaces, AA , and needs modern hardware. But it's still awesome. Handles bump/spec too, sorry pior.
An interesting paper on fast calculation of indirect lighting, could be a similar method to this. Or might be completely different, but still interesting
indirect lighting takes the scene's geometry into account, whilst deferred shading can only lit what is "visible" in the current frame.
With deferred shading the main benefit is indeed that you can do multiple lights on screen at much lower costs. The major drawback is that you only have a fixed set of input paramemters to the shading (normal, position, speccolor, diffusecolor..) that are the same for all surfaces. And as all attributes can only be stored once per pixel, there is no complex transparency, ie transparent surfaces have to be drawn in simpler fashion.
Which means you are very limited in the way you shade things, you dont have as much freedom in different shaders.
Crysis does not use deferred shading in the classic sense, but they use a hybrid approach, certain things are done in screenspace (shadows are collected and applied later), but a lot is still done in regular rendering, where those screenspace maps are used. But the information is not applied like in classic deferred shading, as seen in killzone2, or tabula rasa.
indirect lighting is a different story, as the transport of light is computed for a scene. The fact that light can bounce around means the geometry that is not visible has to be used as well. Ie that is very different from def. shading.
it will take a while until complex material setups as seen in ue3/crysis can go along with techniques presented here. But it's just a matter of time and I like the soft touch to that enlighten stuff. Though I also would be interested in real-life performance, which has been the number one issue for these realtime radiosity engines since the first days they popped up.
Im impressed, I think it looked really great, I was just wondering how it would perform with say 2 large tanks and 12 soldiers running around that place, would they all be lit perfectly too? or would it take a large performance hit to light complex animated vehicles and characters?
MoP, cool paper. Looks different though, they only handle static geom, whereas Geomerics seems to recalculate the indirect for characters too.
Yeah Peris it does seem so, better performance than traditional multipass lighting. The killzone2 and stalker papers both go into detail about advantages/disavantages of deferred shading.
Oh and I remembered STALKER did have helicopters flying around, so there were moving vehicles. But they were so high and fast though that it wasn't clear if they were lit the same way as the rest.
but STALKER doesnt use indirect lighting like proposed by those geomerics, afaik noone does... lightsprint (also realtime radiosity) was said to be licensed for that "film noir" looking Rockstar game. But again in massive environments with lots of changing and moving geometry, there hasn't been a proper realtime game/demo yet.
hence my hint about not mixing up those techniques, that realtime radiosity stuff is different kind of technology. Whether the shading is "deferred" or not is another decision.
Good point. Yeah, in the stalker paper, he talks about how they simulated bounce lighting with a number of performance-optimized non-shadow-casting diffuse-term-only lights. dashity-dash-dash.
the sales blurb on the "what is radiosity page" says this about performance:
The phrase "real-time radiosity" is both over-used and abused in the graphics literature. For use in a game, computing radiosity at 10 frames per second (fps) is simply not good enough. For a lighting technique to be practical then in isolation it must be hitting well over 100fps, so that when it is incorporated into a complete solution, rates of 60fps can be achieved. Enlighten easily fulfils these performance requirements.
Replies
'Real-time' radiosity however, yeesh. Now please?
Neither is the post proccess AO....... I don't think. Actually i am unsure.
Really impressive stuff. Like hellobuddy mentioned, I would be interested to see how this handles in large open areas. But the tech is awesome.
-caseyjones
Exciting stuff
Now show me a demo of a modern building interior with lots of shiny bumpy things using that lighting. If they can do that then they will have allot more sales. As is it looks like a dynamically lit Prince of Persia game.
Anyways the future is bright for artists, no more baking lightmaps and creating lightmap uvs . Deferred renderers also work great with SSAO.
Howmany games are out today that support deferred rendering? I only know of Crysis and GRAW&GRAW2 for pc. I think stalker has one too but i'm not sure.
Thegodzero: normalmaps arnt a problem with a deferred renderer, they actually look alot better becuase alot more specular points at thesame time are possible.
Some papers with nice images...
Killzone 2 (8MB PDF).
http://www.guerrilla-games.com/publications/dr_kz2_rsx_dev07.pdf
STALKER (8MB PDF).
http://www.4a-games.com/209_gems2_ch09.pdf
http://www.tml.tkk.fi/~janne/wrt/
An interesting paper on fast calculation of indirect lighting, could be a similar method to this. Or might be completely different, but still interesting
indirect lighting takes the scene's geometry into account, whilst deferred shading can only lit what is "visible" in the current frame.
With deferred shading the main benefit is indeed that you can do multiple lights on screen at much lower costs. The major drawback is that you only have a fixed set of input paramemters to the shading (normal, position, speccolor, diffusecolor..) that are the same for all surfaces. And as all attributes can only be stored once per pixel, there is no complex transparency, ie transparent surfaces have to be drawn in simpler fashion.
Which means you are very limited in the way you shade things, you dont have as much freedom in different shaders.
Crysis does not use deferred shading in the classic sense, but they use a hybrid approach, certain things are done in screenspace (shadows are collected and applied later), but a lot is still done in regular rendering, where those screenspace maps are used. But the information is not applied like in classic deferred shading, as seen in killzone2, or tabula rasa.
indirect lighting is a different story, as the transport of light is computed for a scene. The fact that light can bounce around means the geometry that is not visible has to be used as well. Ie that is very different from def. shading.
it will take a while until complex material setups as seen in ue3/crysis can go along with techniques presented here. But it's just a matter of time and I like the soft touch to that enlighten stuff. Though I also would be interested in real-life performance, which has been the number one issue for these realtime radiosity engines since the first days they popped up.
Yeah Peris it does seem so, better performance than traditional multipass lighting. The killzone2 and stalker papers both go into detail about advantages/disavantages of deferred shading.
Oh and I remembered STALKER did have helicopters flying around, so there were moving vehicles. But they were so high and fast though that it wasn't clear if they were lit the same way as the rest.
hence my hint about not mixing up those techniques, that realtime radiosity stuff is different kind of technology. Whether the shading is "deferred" or not is another decision.
The phrase "real-time radiosity" is both over-used and abused in the graphics literature. For use in a game, computing radiosity at 10 frames per second (fps) is simply not good enough. For a lighting technique to be practical then in isolation it must be hitting well over 100fps, so that when it is incorporated into a complete solution, rates of 60fps can be achieved. Enlighten easily fulfils these performance requirements.