ok then, but intentionally or not you did make a particularly statement, namely that UMA with massive badnwidth would be some sort of a panacea if it wasn't for its cost. well, that's not true. it's not the cost, it's the extra non-deterministic latency that UMA introduces, which, firstly, has nothing to do with bandwidth, and secondly, is outright bad(tm) in the context of modern cpu architectures.
Rumble Roses developer on PS3 & X360
- Thread starter seismologist
- Start date
Black Dragon37
Newcomer
And that's their flaw - their reluctance of technology change.onetimeposter said:
Shifty Geezer said:
The problem of completely simulating human motion has not been solved yet to a level that would be acceptable to the general audience. There's been procedural, footstep-based animation in character studio for 3ds max since 1996 or something, but it always looked horrible.
So while you may have something that looks right from a scientific viewpoint, it would still lack the emotion, the acting, that is added by either the mocap actor or the animator. You want an enemy to look menacing, powerful, you want your hero to sneak? All those things are the result of small touches that sometimes even have to be unrealistic to sell the animation to the audience. It's why we call both acting and animation an art, and science has not yet managed to proove that it can match the skills of a talented artist.
It is very evident to us in the production of CGI that as you start to apply algorithmic corrections to mocap data, it will start to loose it's realism, it's character, it's extra touch. This is why motion editing is usually performed by humans as well, and it's better to have guys with animation experience and talent to get good results. The computer still does not know how a confident walk cycle looks... unless you want to do robots only, and even those robots would lack any character.
The short answer is that you shouldn't expect more than the ragdoll physics that you've seen so far. Maybe fitting the feet on the ground with IK as well, but that's all - the animation cycles should still be created with mocap or keyframing.
Facial animation is even more reliant on artistic talents, in order to express emotions and thoughts. Thus, forcing the animator to think in muscle movements instead of expressions would limit their abilities - although this could be improved to a level by abstracting the controls (hiding the individual muscle movements by grouping them into expressions). But the bigger problem is that simulating something is always much more complicated and time consuming then manually creating the desired result.
Also, a simulation model is usually very slow, certainly slower than it's alternatives. You want to have it running at an interactive framerate, partially to allow fast feedback to the animator and partially because it should run in the game
To give an indication, the creatures in Hellboy used a muscle simulation system for the skinning of the body. Running a 10-30 second shot's simulation usually required a whole night's calculations on a large renderfarm of PCs. You'd need at least two orders of magnitudes of increase in computing power to be able to simulate just one of these creatures in real time.Some game devs tend to mention that they're doing some kind of muscle simulation. In reality it's that they add a few extra bones into a character rig that try to act as muscles, with their movement manually tweaked by an artist to help with the trickier deformations.
For example, you could add an extra bone to 'simulate' the secondary movement of a woman's breast, but we all know that it doesn't have anything to do with true simulation as there aren't any bones in there
Shifty Geezer said:
The problem is that for character animation, there aren't really any better options than to either mocap or manually animate. The movie VFX industry would be more than happy to spare the money spent on these if there would be cheaper and better looking options but so far, there aren't.
That's why they mocap hundreds of motions for crowd scenes in movies like Troy or LOTR, which then get cut to tiny parts that can be re-assembled into whatever the crowd simulation requires. Games don't have the budgets to do this, so the animation blending won't be as good.
You paint quite a disappointedly pessimistic view of next-gen graphics
If the processing requirements (ignoring need for research into realistic and natural motion) are astronomically beyond anything a good few generations of consoles are going to be able to do, that would suggest that for a LONG time all animations are still going to be mo-caps, for faces and skeletons. Sword swinging is still going to be mo-cap based, so there won't be realistic responses based on the material being hit or an enemy parrying your strike. Characters with walk cycles won't be able to adapt to terrain and walk over obstacles.
I'll be disappointed at this. The idea that animation won't improve much save for more motion caps for the next several generations of tech means really, what are we getting other than improved graphics? And based on your other points for graphics, like not enough power for any useful raytracing in graphics or what have you, it seems you feel all that floating point power isn't going to be very 'useful'. It's not going to be doing anything next-generationy regards the technologies used to create graphics and games that we already have. As a personal POV what do you think the floating point power of next-gen and subsequent generations of consoles which actually bring to gaming visuals? If the underlying techs are no different to what we have now, are you expecting nothing other then what we have now, but with more characters and more barrels and more shaders? Do you not think there'll be hybrid developments and new techs to achieve more versatile and realistic animations?
If the processing requirements (ignoring need for research into realistic and natural motion) are astronomically beyond anything a good few generations of consoles are going to be able to do, that would suggest that for a LONG time all animations are still going to be mo-caps, for faces and skeletons. Sword swinging is still going to be mo-cap based, so there won't be realistic responses based on the material being hit or an enemy parrying your strike. Characters with walk cycles won't be able to adapt to terrain and walk over obstacles.
I'll be disappointed at this. The idea that animation won't improve much save for more motion caps for the next several generations of tech means really, what are we getting other than improved graphics? And based on your other points for graphics, like not enough power for any useful raytracing in graphics or what have you, it seems you feel all that floating point power isn't going to be very 'useful'. It's not going to be doing anything next-generationy regards the technologies used to create graphics and games that we already have. As a personal POV what do you think the floating point power of next-gen and subsequent generations of consoles which actually bring to gaming visuals? If the underlying techs are no different to what we have now, are you expecting nothing other then what we have now, but with more characters and more barrels and more shaders? Do you not think there'll be hybrid developments and new techs to achieve more versatile and realistic animations?
I think that some of the problems you see in the character animation are inherent. For example you can't have a player controled character to react to the controls immediately and in a realistic way. If changing the game mechanics, the 'feel' of the controls is not an option than you can't do anything about it.
Also, some cases like moving over a small obstacle require the animation engine to know about the problem before it appears, ie. realize that there will be a collision before it happens. That's how it goes in reality, you see the obstacle and will move your leg differently as you start your next step. I guess it could be solved with clever programing but some of the problem remains - what if the character starts to move over the obstacle, but the player suddenly makes it stop? It could probably look just as silly. AI characters are somewhat easier because the animation engine can ask the AI about what it plans to do, but the player is unpredictable.
Nevertheless, I'm sure that there's room for some extent of procedural animation in characters. But this will be more of an extension of the currently used technologies than a replacement.
Simulations will be implemented to an extent... not too complicated cloth and cloth-based hair dynamics are already there in Heavenly Sword and even in MGS. We can expect some more of that, but certainly not something as complicated as Yoda's robe, for example. Muscle simulation (and more importantly, skin simulation) will remain too expensive for this generation IMHO. As long as it takes hours and not minutes to calculate it offline, I would not expect it to work in real time on the consoles, that are only about an order of magnitude faster in FP calculations than regular PCs.
I think the most important feature that this gen is going to bring is realistic dynamic lighting and shadows. What we've had on the previos gen wasn't too good, don't you think so? Thus I believe that you're expecting a bit too much
Also, there seem to be some really large draw distances in quite a few games, and having more than a dozen characters on screen is going to add a lot as well.
Also, some cases like moving over a small obstacle require the animation engine to know about the problem before it appears, ie. realize that there will be a collision before it happens. That's how it goes in reality, you see the obstacle and will move your leg differently as you start your next step. I guess it could be solved with clever programing but some of the problem remains - what if the character starts to move over the obstacle, but the player suddenly makes it stop? It could probably look just as silly. AI characters are somewhat easier because the animation engine can ask the AI about what it plans to do, but the player is unpredictable.
Nevertheless, I'm sure that there's room for some extent of procedural animation in characters. But this will be more of an extension of the currently used technologies than a replacement.
Simulations will be implemented to an extent... not too complicated cloth and cloth-based hair dynamics are already there in Heavenly Sword and even in MGS. We can expect some more of that, but certainly not something as complicated as Yoda's robe, for example. Muscle simulation (and more importantly, skin simulation) will remain too expensive for this generation IMHO. As long as it takes hours and not minutes to calculate it offline, I would not expect it to work in real time on the consoles, that are only about an order of magnitude faster in FP calculations than regular PCs.
I think the most important feature that this gen is going to bring is realistic dynamic lighting and shadows. What we've had on the previos gen wasn't too good, don't you think so? Thus I believe that you're expecting a bit too much
Also, there seem to be some really large draw distances in quite a few games, and having more than a dozen characters on screen is going to add a lot as well.
still ,there is lot of room for tricks.
You can do realtime muscle (-like ) simulations with hierarchical deformers ,lattices , driven control ,expressions ,and so long.
Same goes with animation.Dynamic physics-based or procedural modifications,more complex IK solving with some retargetting will be possible.
You can do realtime muscle (-like ) simulations with hierarchical deformers ,lattices , driven control ,expressions ,and so long.
Same goes with animation.Dynamic physics-based or procedural modifications,more complex IK solving with some retargetting will be possible.
Yes, this sounds about right. Lighting and shadows are an important step forward, though this seem more GPU based then CPU. I do hope some of that float power can be used in animation situations though (not that that'll help in the screenshot-wars!)Laa-Yosh said:I think the most important feature that this gen is going to bring is realistic dynamic lighting and shadows. What we've had on the previos gen wasn't too good, don't you think so? Thus I believe that you're expecting a bit too much
Also, there seem to be some really large draw distances in quite a few games, and having more than a dozen characters on screen is going to add a lot as well.
_phil_ said:
That is not a simulation in the industry's terms. These are common tricks in offline CGI, I've used them many times as well, but it's got nothing to do with simulation. Just as we won't call a piece of clothing moved by sime bones cloth simulation, we won't call skinning tricks a muscle simulation either.
Real muscle simulation may involve calculating the muscle volume, muscle-muscle and muscle-bone collisions, softbody dynamics and even gravity. More importantly, it will have to calculate the way the skin wraps over the muscle and optionally the fat tissues - which is generally very similar to cloth simulation. All in all it involves solving a formula for each vertex that probably involves physics-related calculations.
I know all that (started Cg with 3ds 3 ) .In my words ,simulation means visual imitation.Most of realtime stuff is about tricks ,and a lot of heavy offline CG stuf can be more or less convincingly approached with tricks.
Why waste processing power simulating (in 'industry's terminology') things if you can make it behave or look good in an interactive environement.Outside of PR blabla hype ,there is realy no point.
btw ,check GPGPU.org from time to time , interesting stuff there.
Why waste processing power simulating (in 'industry's terminology') things if you can make it behave or look good in an interactive environement.Outside of PR blabla hype ,there is realy no point.
btw ,check GPGPU.org from time to time , interesting stuff there.
Last edited by a moderator:
How do you make the distinction between fake simulation and real simulation then? The industry prefers to call only the real stuff simulation (while keeping in mind that even the most complicated methods are simplified approximations of the real thing).
I suggest we stick to this convention here on B3D as well.
I suggest we stick to this convention here on B3D as well.
_phil_ said:
Obviously because the fake result will not look like the real one. Getting proper secondary dynamics for the muscles and skin sliding is not possible without a real, full-blown simulation. Even the Hulk movie used skin simulation, although they weren't simulating the muscles (they've sculpted hundreds of blendshapes instead, to get better artistic control).
Same goes for cloth and hair and fluids. Looong offline calculations cost a lot of time, but the results are so much better than what you can do with fake tricks that it's worth doing it for CG. Games don't have the computing power to do simulations so they will stick to tricks - but please don't call these tricks a simulation as well, let's make a distinction here.
MechanizedDeath
Regular
I think you look at this from too much of a CGI pov. Games will not achieve that level of realism. And the mocap stuff is still very unrealistic. I think stuff shown in that Natural Motion site suggests that a good approximation is very much realistic in realtime. I don't see why you would need to have realistic muscle flexing and skinning when a realistic walk/run animation alone would do as much for bipedal character realism as mechanical physics did for racing games.
Simulation, emulation, trepidation...it doesn't matter what it's called. I don't think semantics need apply when it's clear we're talking about realtime applications of physics modeling. I think it's already been shown that some of the other CGI stuff you've talked about in the past has been faked to an acceptable level thus far with next-gen visuals. So we don't need the same level of accuracy for physics-based animation either. PEACE.
Simulation, emulation, trepidation...it doesn't matter what it's called. I don't think semantics need apply when it's clear we're talking about realtime applications of physics modeling. I think it's already been shown that some of the other CGI stuff you've talked about in the past has been faked to an acceptable level thus far with next-gen visuals. So we don't need the same level of accuracy for physics-based animation either. PEACE.
The whole simulation topic was not raised by me. I'm only trying to provide some insight by explaining how these things work in offline CG, which is most of the time an inspiration and example for realtime applications. If you don't like it, skip my posts.
Laa-Yosh said:
this is the whole point.Simulation is not garanted to look better (more often the countrary) .That's true for hair ,same goes for muscles.
What you call real is a mathematical approach of a problem.there is a lot of differents approaches for hair ,for exemple.
They are not less fake.just more heavy on mathematical modeling.
So i don't see why one would be called fake or real .none is real.In fact ,the "real" one have good chances to propel us right in the uncanny valley.
Most of the time ,an approach with more artistical control will be better to achieve correct end user perceptual emotion which is what counts ultimately (in entertainement).
mckmas8808
Legend
Man as long as it looks great and is believable that's all I care about. I like reading about your job and what you do Laa-Yosh, but sometimes I and some others here just wish you stop painting a doom and gloom brush over what next-gen games will and won't have. I think MechanizedDeath has pointed out that you have been wrong some far on one or two things when it comes to the visuals (and this is before any next-gen system comes out) so telling us what's not going to happen with physics is something that we are not going to believe either.
Like this quote from you for example.
Totally false. There is no way that this quote will hold any truth.
Like this quote from you for example.
Totally false. There is no way that this quote will hold any truth.
Hey, don't go! I like hearing your take on matters. Nice to hear the viewpoint from someone in on the graphics side. And I've a question to
How do you think animation will be handled in a simple one on one fighting game like Tekken 6? Do you think Inverse Kinematics can be a major component, or will it still be based on sections of mocap pieced together? I'm thinking as an example of Lee (the Chinese Five Arts Tai Chi Cop) from Tekken...3 or whatever it was. He'd move through several Tai Chi positions with a rigid, unnatural transition between them. What techniques do you think can be used to generate more natural animations in next-gen?
How do you think animation will be handled in a simple one on one fighting game like Tekken 6? Do you think Inverse Kinematics can be a major component, or will it still be based on sections of mocap pieced together? I'm thinking as an example of Lee (the Chinese Five Arts Tai Chi Cop) from Tekken...3 or whatever it was. He'd move through several Tai Chi positions with a rigid, unnatural transition between them. What techniques do you think can be used to generate more natural animations in next-gen?
Tekken is so much more responsive than most other games. Animation may not look natural but it is fluid.
However, the nature of the game isn't terribly realistic. You see things which goes against your intuition, like fighters backflipping away from their opponent but at the same time administering forceful blows. Or how about what appears to be lightweight characters throwing around others which appear to weigh several times their weight?
I guess they have to do it this way or else nobody will choose certain characters. But it's an arcade game, not a fighting simulation. So maybe the animations and motion can be exaggerated in some ways.
However, the nature of the game isn't terribly realistic. You see things which goes against your intuition, like fighters backflipping away from their opponent but at the same time administering forceful blows. Or how about what appears to be lightweight characters throwing around others which appear to weigh several times their weight?
I guess they have to do it this way or else nobody will choose certain characters. But it's an arcade game, not a fighting simulation. So maybe the animations and motion can be exaggerated in some ways.