Shadow of the colossus and framerate issues

london-boy said:
Even realtime raytracing is faking reality.
Only reality is reality.
Even if something looks completely real, as long as we have to process it and see it through a display device like a TV or a monitor (even a 3D monitor), it's still fake.

But will you choose the red pill, or the blue pill?
 
MrWibble said:
The nearest casualty department, I shouldn't wonder. :)

Speaking of strange substances, am i the only one who for a moment wondered what kind of drugs Ico's director was on when he thought about both Ico and SOTC? Especially SOTC.
 
london-boy said:
Even realtime raytracing is faking reality.
Only reality is reality.
Even if something looks completely real, as long as we have to process it and see it through a display device like a TV or a monitor (even a 3D monitor), it's still fake.
A movie camera captures reality. Though people frame reality to look like something it isn't, if you just set up a camera and start filming, what it records is real. What it shows ont he TV is real. Now we can argue that the image doesn't exist as a reality etc. But the way cameras work to record reality to is sample points for photons bouncing off objects, the same way our eyes so when looking at a real scene rather than a projection of a captured real scene. When CG's can produce in the same quality a virtual world with the same movement of photns and a virtual camera that records photon samples to show those measurements as dots that make up an image, then we're creating a rendering of a world without faking it, but creating images just how it is in the real world. If that makes sense. Fur is lots of thin flexible cylinders that reflect light. Where at the moment we draw hair in games by 2D textures on planes, in the future when we can instead render fur as millions of thin flexible cylinders reflecting light, we're no longer faking the rendering of hair, but simulating it. If that makes sense.

Of course, this renders critical philosophical questions. If the world is created based on the same laws as this world, does the world ACTUALLY get created? When you switch off the console are you destroying a world and all it's inhabitants? Will game characters have a soul, and will a virtual crime against virtual people actually be as morally wrong as it is in the real world against real people?

I think I ought to get in early. I'll create a legal firm to represent virtual computer characters. All the mushrooms Mario stomps on deserve legal representation! They've rights too, ya know!
 
Shifty Geezer said:
A movie camera captures reality. Though people frame reality to look like something it isn't, if you just set up a camera and start filming, what it records is real. What it shows ont he TV is real. Now we can argue that the image doesn't exist as a reality etc. But the way cameras work to record reality to is sample points for photons bouncing off objects, the same way our eyes so when looking at a real scene rather than a projection of a captured real scene. When CG's can produce in the same quality a virtual world with the same movement of photns and a virtual camera that records photon samples to show those measurements as dots that make up an image, then we're creating a rendering of a world without faking it, but creating images just how it is in the real world. If that makes sense. Fur is lots of thin flexible cylinders that reflect light. Where at the moment we draw hair in games by 2D textures on planes, in the future when we can instead render fur as millions of thin flexible cylinders reflecting light, we're no longer faking the rendering of hair, but simulating it. If that makes sense.

Of course, this renders critical philosophical questions. If the world is created based on the same laws as this world, does the world ACTUALLY get created? When you switch off the console are you destroying a world and all it's inhabitants? Will game characters have a soul, and will a virtual crime against virtual people actually be as morally wrong as it is in the real world against real people?

I think I ought to get in early. I'll create a legal firm to represent virtual computer characters. All the mushrooms Mario stomps on deserve legal representation! They've rights too, ya know!


Well i think you're confusing a "perfect" fake with today's fake.
As long as we process data to get a certain look, it's a fake. The fact that it might look 110% real just means that it's a VERY GOOD fake. Still a fake.
A camera filming reality is not a fake, it's a representation of reality. Could be considered fake by some people, personally i don't.
CGI, as good as it can get in the future, even if it looks indistunguishable from reality, with 3D displays and Virtual Reality stuff, will ALWAYS be a fake.
 
london-boy said:
Well i think you're confusing a "perfect" fake with today's fake.
As long as we process data to get a certain look, it's a fake. The fact that it might look 110% real just means that it's a VERY GOOD fake. Still a fake.
A camera filming reality is not a fake, it's a representation of reality. Could be considered fake by some people, personally i don't.
CGI, as good as it can get in the future, even if it looks indistunguishable from reality, with 3D displays and Virtual Reality stuff, will ALWAYS be a fake.

Hmmm.

What is the important difference between:

Taking a limited number of discreet samples of light that hit a surface at a particular point in a physical scene which can be reproduced to approximate what the geometry looks like to an observer.

Taking a continuous representation of the geometry of a physical scene and then simulating (very accurately) the light transport in the scene to approximate what the scene looks like to an observer.

?

If the position and number of samples used in the former are applied in the latter and give the same result, why is one fake and the other not?

If a tree falls in the forest and there is no-one there to hear it, is there any point rendering it?
 
Last edited by a moderator:
MrWibble said:
Hmmm.

What is the important difference between:

Taking a limited number of discreet samples of light that hit a surface at a particular point in a physical scene which can be reproduced to approximate what the geometry looks like to an observer.

Taking a continuous representation of the geometry of a physical scene and then simulating (very accurately) the light transport in the scene to approximate what the scene looks like to an observer.

?

If the position and number of samples used in the former and applied in the latter, and give the same result, why is one fake and the other not?

If a tree falls in the forest and there is no-one there to hear it, is there any point rendering it?

Well i think this whole idea is quite subjective, but i think that "making" a representation of reality, even with infinite detail which no one will ever distinguish from reality, is a fake. A fake as in a representation of reality, not the real thing.
Filming something, dunno i think i could consider that as a fake too, a representation of reality.

Only reality is real, everything else is a fake.

Does that make me a cynical bastard?
 
london-boy said:
Well i think this whole idea is quite subjective, but i think that "making" a representation of reality, even with infinite detail which no one will ever distinguish from reality, is a fake. A fake as in a representation of reality, not the real thing.
Filming something, dunno i think i could consider that as a fake too, a representation of reality.

Only reality is real, everything else is a fake.

Does that make me a cynical bastard?

Yes!:LOL: Are you guys hitting the bong this morning? Getting kina deep in here.
 
london-boy said:
Well i think this whole idea is quite subjective, but i think that "making" a representation of reality, even with infinite detail which no one will ever distinguish from reality, is a fake. A fake as in a representation of reality, not the real thing.
Filming something, dunno i think i could consider that as a fake too, a representation of reality.

Only reality is real, everything else is a fake.

Does that make me a cynical bastard?

Ah, grasshopper... you have only taken the first step on the road to being a truly cynical bastard...

I think the view that reality is real and everything else is "fake" by virtue of being only an approximation would be at least consistent.

If you're going to qualify some of the approximations (such as film or photo) as also being "real" in some sense, then you have to extend that definition to a suitably advanced reproduction provided the simulation was also based on real-world data and wasn't making something up.

Er.. Shadow of the Colossus... it's really great, isn't it?
 
mckmas8808 said:
Yes!:LOL: Are you guys hitting the bong this morning? Getting kina deep in here.

Well apparently l-b is having some kind of puppy love craziness (judging by the thread over in the general forum).

No idea what my excuse is...
 
london-boy said:
Well i think this whole idea is quite subjective, but i think that "making" a representation of reality, even with infinite detail which no one will ever distinguish from reality, is a fake. A fake as in a representation of reality, not the real thing.
In terms of CG's though, we're never going to get better than simulation real-world light-transport rendering. Unless we actually create matter Holodeck style. If we take simulating light transport realistic as a 'real' solution to hair rendering, then anything that renders hair differently is a 'fake'.
 
london-boy said:
As has been said many times, it's not exactly "Fur Shading" the way it's done with pixel shaders on PC or Xbox. It's "fake", but in the end it looks like fur so one can still call it "Fur Shading". The fact that it's done by overlaying LOTS of alpha textured polygons (bit like smoke effects, but fixed to the Colossus surface and with a "hair" texture on them) shouldn't detract it from being called "Fur".
Cool. so You're saying that the colossus are made with lots of polys? If so no wonder it's poly count was a little low. dont get me wrong, it was enough polys make it look good, but things were a little blocky. And i always thought particles always faced the screen. so anyway what's missing to make the ps2 version of "fur-shading". and how many passes will it take on ps2 to do xbox's fur shading?
 
Did they really make the fur just by using lots of polys? That doesn't seem very practical to me, because when you look up close there seems to be at least 7 or so "layers" to the fur. The polygon count would be astronomical. Still, however they did it, it looks amazing.
 
Fur

Branduil said:
Did they really make the fur just by using lots of polys? That doesn't seem very practical to me, because when you look up close there seems to be at least 7 or so "layers" to the fur. The polygon count would be astronomical. Still, however they did it, it looks amazing.

This is a common fur modeling technique called fin and shell. It is a good model that allows rotation around furry object and for light to filter through fur on object's silhouette. Other models available, such as texture method used for grass depiction in Spider-man games can be used to reduce load but will not be effective where very close camera view is required. Also available is 3D texture technique which is most effective and does not have dotted look of SOTC or Conker and I believe Nintendo games might have used this method.

Interesting article on 3D texture method:
http://www-graphics.stanford.edu/courses/cs348b-competition/cs348b-03/fur

Because SOTC is designed for very close camera view and uses fin and shell model, fur density has to be very high to reduce dotted appearance. I do not know precise polygon count but surely is very high because of extent of fur simulation. This is necessary because long time duration of very very close camera view must be accomodated. However, clever LOD system reduces load.
 
Branduil said:
Did they really make the fur just by using lots of polys? That doesn't seem very practical to me, because when you look up close there seems to be at least 7 or so "layers" to the fur. The polygon count would be astronomical. Still, however they did it, it looks amazing.

Well the polygon count could get quite high, but not astronomical.
 
london-boy said:
Well the polygon count could get quite high, but not astronomical.

For the furry patches you might be looking at 10x the number of polygons and fillrate for doing 10 layers. That's probably in the ball-park for the games discussed here.

Given that there is only one (albeit huge!) character on screen using fur, and even then only in patches, the number of polygons is probably quite manageable. Especially as they're trivially transformed from the base polygons and so cheap to calculate. The fillrate is no problem for the PS2, so on the whole it's not too hard a technique. You can probably write an engine churning out more polygons per frame using stuff like that (or any multipass technique) than if you try to use the extra polygons to make more detail or add more instances of characters.

Thats for the shells - if you add fins then it gets more interesting because they look bad (and can make the whole fur effect look worse) unless you blend them in well at the right viewing angles, and thats not necessarily cheap. For those though, you're only adding a small number of polys. At most, with the nastiest available hack, you're adding one quad per edge, which is in the order of one more textured shell.

However it's actually surprisingly hard to make it look really good, so you need to experiment a lot with the textures involved. The SotC guys did a nice job. If I was highly unscrupulous I'd be very tempted to poke around and see exactly what their rendering passes look like.
 
MrWibble said:
For the furry patches you might be looking at 10x the number of polygons and fillrate for doing 10 layers. That's probably in the ball-park for the games discussed here.

Given that there is only one (albeit huge!) character on screen using fur, and even then only in patches, the number of polygons is probably quite manageable. Especially as they're trivially transformed from the base polygons and so cheap to calculate. The fillrate is no problem for the PS2, so on the whole it's not too hard a technique. You can probably write an engine churning out more polygons per frame using stuff like that (or any multipass technique) than if you try to use the extra polygons to make more detail or add more instances of characters.

Thats for the shells - if you add fins then it gets more interesting because they look bad (and can make the whole fur effect look worse) unless you blend them in well at the right viewing angles, and thats not necessarily cheap. For those though, you're only adding a small number of polys. At most, with the nastiest available hack, you're adding one quad per edge, which is in the order of one more textured shell.

However it's actually surprisingly hard to make it look really good, so you need to experiment a lot with the textures involved. The SotC guys did a nice job. If I was highly unscrupulous I'd be very tempted to poke around and see exactly what their rendering passes look like.

Definately interesting to see how they managed to make it look so good.

Not sure this has been covered in the thread, but i'm pretty sure the Colossi animation employs inverse kinematics, from the way they move and especially the way they fall to the ground. Very realistic and it just looks like inverse kinematics, although i still have to play the game and only saw one Colossus fall to the ground (from the demo videos).

Regardless, i think we're starting to see why SOTC looks the way it looks and performs the way it performs. Those Colossi not only are bloody huge, but have so much going for them that it's easy to see why slowdowns occurr on little old PS2, which still amazes me, considering it's almost 6 years old. then again, when one team has the luxury to take 3 years off (fully paid by Mr Kutaragi) and focus on one single game for one single platform, i guess they can pull tricks other people can't.
 
london-boy said:
Not sure this has been covered in the thread, but i'm pretty sure the Colossi animation employs inverse kinematics, from the way they move and especially the way they fall to the ground. Very realistic and it just looks like inverse kinematics, although i still have to play the game and only saw one Colossus fall to the ground (from the demo videos).

I'd say there is almost certainly IK in places, though maybe not in the situation you mention. I think the death sequences, to me, look like canned animation.

However there are other motions that probably involve IK in one form or another.

Still no sign of your multi-region thing?
 
Back
Top