Next generation animation

[wco81]

I don't think anyone is against mocap per se.

Just that next-gen games show more dynamic animations reflecting more realistic physical modeling as well as variety.

Whether that's achieved by mocap or some entirely different system or some synthesis of mocap and something else would be irrelevant.

It looks like they cooked up that football tackling demo to lobby for EA's business. The company is European right? There's only one NFL game now and the demo of American football animation can't be of much use to Euro developers.

Would be nice to see in soccer games though.

 

[Alejux]

Fully believable synthesized active motion hasn't even been done in research yet. It's just too complex for today's hardware, let alone for realtime. Everything we see today uses fragments of mocap or carefully scripted animations jumbled around.

 

[Laa-Yosh]

Fully believable synthesized active motion hasn't even been done in research yet. It's just too complex for today's hardware, let alone for realtime. Everything we see today uses fragments of mocap or carefully scripted animations jumbled around.

It's not just that it's too complex for the hardware, it's that the problem hasn't even been solved yet

 

[pc999]

I think that we will see many new animation modes, thinghs are geting (eg.terrein) very complex, which means that there is a lot of variables and usually motion capture are too limit (that why in the game we see them walking on the edge with on2 feet on air), those like Endorphin (even using variations/variating of/the motcap).

In a second thought I think that Rev will present a true challenge to animations (especially in TPS or MP games), what hapens if the player cut (by bad luck) is leg in a sword game, will the sword pass by it like if as nothing, will not pass, will make a little damage/cut it all, or will he jump, what hapens will be a gameplay decission but most of the options looks like hard job to do in the animation side (that is the only thing that make me think if there is really any change too get a mini PPU or something like that in REV, once that you will probably need some very acurrate physics to a game like this and it would help a lot in animations).

Interesting times head I think.

 

[mckmas8808]

You guys do realise that these Endorphin animations are built from standard mocap running and jumping sequences, right? Endorphin takes care of the collision and what happens after it, until the next canned animation starts - but the rest of the animation still has to be created somehow.

Yep I know that and I hope EA using it too. That clip of Endorphin is damn near perfect. Just two thing need to happen to make it perfect.

1. The tackler needs to wrap up the runner with the ball.

2. The runner obviously can't let the ball go each and everytime he gets hit.

 

[EricVonZipper]

http://www.gametrailers.com/gamepage.php?id=2199

These Zombies wrap up the 'runner' pretty well! This really looks good and I hope people can appreciate whats happening in this short clip.

I've watched my fair share of hands, feet , ect..., passing thru walls, players, objects and to see this level of clipping detect is very cool.

The fanites at teamxbox had a hard time understanding what this clip was showing them.

Not sure if any system will be able to handle this on a large scale.

http://www.worthplaying.com/article.php?sid=28750&mode=thread&order=0


Chris Swan, Project Manager of Possession for Volatile Games enthused, "Incorporating endorphin 2.5 into the Possession animation pipeline has been one of those rare experiences where you receive a significant increase in quality and a saving in time all in one go!" He continued, "Our next-gen title relies heavily on our zombies and civilians attacking each other in very detailed and original ways. There's only so far that you can go with real-life actors of course, which is where endorphin steps in. In minutes we can literally load in some motion capture footage, set up a few behaviours on the characters, and let endorphin work out the rest! The resultant animation is incredibly convincing (we haven't had to touch up a single animation so far), while the save in hand-animation time is huge."

 

[Qroach]

natural Motion is really cool, and it is an extremly powerful animation tool. I can see it being a great use when you want to mix mocap with physical animaiton to create new animations for a game, but:

1. This won't work very well in a realtime game where you don't have the luxry of someone setting up contraints dynamically to make a scene look right. If you watch the demonstration videos you'll quickly realize that to get the best effect, they use key frames an mocap data, constantly switch between pre made data and dynamic data. There's no way you can do that in a sports game because you simply don't have the luxry of knowing what's going to hit what far enough ahead of time.

 

[AlNom]

I've watched my fair share of hands, feet , ect..., passing thru walls, players, objects and to see this level of clipping detect is very cool.

Why not just implement per polygon collision detection ala Doom 3

 

[EricVonZipper]

Why not just implement per polygon collision detection ala Doom 3

High detail models are making accurate collision detect incredibly difficult, most high poly objects are just bounding volumes. I get a headache just thinking of the sub-routines to detect say..intertwining fingers....even a handshake is VERY difficult.(probably some sort of fluid technology)

I look forward to the CryENGINE 2 with the collision detect on plants, that is amazing stuff!

Do a Google search for Collision detection if you want a headache.

And I'm pretty sure Doom3 uses a 'quick pass bounding box' most of the time.

Ughh I hate CD.

http://graphics.ethz.ch/~brunoh/download/VolumetricCollisionDetection_TR03.pdf

 

[AlNom]

And I'm pretty sure Doom3 uses a 'quick pass bounding box' most of the time.
[/url]

But isn't that good enough? I don't ever really notice clipping, and performance only bogs down when there are actually a lot of collisions going on.

 

[EricVonZipper]

But isn't that good enough? I don't ever really notice clipping, and performance only bogs down when there are actually a lot of collisions going on.

Yes for first person shooters I think it's fine.

I think D3 weapon system identifies per poly collision, so it can identify head, torso, appendage shots. A creature shambling down a corridor requires little detection.....just enough to keep him from getting stuck in a wall.

Wrestling games, fighting games still suffer from hands, feet, head, ect... passing thru objects.

I know of no realtime engine(besides cryENGINE2)that does accurate collision detect.