Clothes in 3D computer graphic

[Gian_Yagami]

Hi, I am just curious that video games nowadays looks hyper realistic, clothes and body looked separately, how could they rig clothes model and naked body together? in example gta v, clothes have folds and it changes dynamically according to posture.

[Tchuck]

Usually done by adding a lot of bones on the clothes and animating them together with the body. I remember working on FFXV, the model for Noctis was very detailed, and the clothing he wore had a lot of bones added to it for certain detailed areas. Most of the clothing mesh was skinned with the regular body bones; some detailed areas (like things hanging etc) had special bones.

 

The rigs we had were pretty high level, affecting the movement of the clothing as you moved an arm/leg etc for animating.

 

For cutscenes etc where we could use higher polygon models for more detail, we would usually simulate the cloth using physics, making them look even more realistic with all the folds and whatnot.

[Gummy]

I'm pretty curious about this as well. Saw the detail on some of the characters in Uncharted 4 and it blew my mind.

[Dryspace]

@Tchuck

 

You are saying that loose and hanging parts of a character are not usually responding to gravity in real time, but are just animations tied to the character's animation? If so, I did not know that.

 

There is a demo called Nurien Alpha (my copy is named nzd_Nurien_Alpha_v0.7.exe) that demonstrates real-time cloth physics in the form of a woman dancing in a dress. I believe it came out right around the abandonment of the PC AAA industry.

[Tchuck]

 

You are saying that loose and hanging parts of a character are not usually responding to gravity in real time, but are just animations tied to the character's animation? If so, I did not know that.

 

I'd say in 99% of the cases, correct. Physics is an expensive thing to calculate, and even though we have an ever increasing amount of power at our disposal, we wanna be efficient. We cut a lot of corners in game development. If we can achieve a similar looking effect, but cheaper, we will do it. We could use physics for all of those things, but a decent rig will take care of animating them under most circumstances. Beyond rigs, there's also systems that will affect other joints based on specific functions/parameters, so that when the leg raises, the skirt automatically moves around to fit it etc.

[Gian_Yagami]

I thought character model and clothes model are separately. Just like putting clothes into a mannequin. For games which have clothing system like the sims or gta v

[Tchuck]

They are modeled separately. Most clothing objects will share the same skeleton, though. After being modeled in a T-pose, it is bound to the skeleton, weighted, and then tested with the stock animations. If a specific item of clothing doesn't make use of features other clothing does (say, Jacket A has a lot of blingy things, Jacket B doesn't) they still use the same skeleton, the same animations, but the game is either structured to know when to ignore movement to specific bones if they aren't being used, or they are animated anyway and since no vertex is bound to them, nothing really happens.

[Gian_Yagami]

They are modeled separately. Most clothing objects will share the same skeleton, though. After being modeled in a T-pose, it is bound to the skeleton, weighted, and then tested with the stock animations. If a specific item of clothing doesn't make use of features other clothing does (say, Jacket A has a lot of blingy things, Jacket B doesn't) they still use the same skeleton, the same animations, but the game is either structured to know when to ignore movement to specific bones if they aren't being used, or they are animated anyway and since no vertex is bound to them, nothing really happens.

I got it about extra bone, just like max payne 2, max have jacket which always moving dynamic. And about separately what I mean like this, is it correct?

Old games like gta san andreas using body piece (i.e T-shirt are modeled with hand and it's modeled only for CJ). What about gta v which have three protagonist + character creation.

 

Anyway I am glad a developer like you had replying my thread.

[Tchuck]

So, all humanoids tend to share the same skeleton. The skeleton can be adjusted for special cases, sometimes dynamically during runtime, but all animations that can be used on one skeleton, can easily be retargetted for another skeleton: the reason being that most animations will only affect the rotation attribute of the bones. This way, even if one character's arms is longer than another's, the animation can still be used for either of them.

 

As for modelling all them separately, it depends.

If your game features clothes change, and we see the character down to his undies, then we model the entire body, and model the clothes on top for better reference. There's other tools for modeling clothes, but I can't go into deep detail on them.

If you game features clothes change, and we never really see the character without clothes, then we model characters heads and arms, and then the clothes.

If there's no clothes change, we just model the character as they will look like.

[Dryspace]

Are most characters still animated using only key-frame or motion capture, or are real-time algorithms commonly used which modify "canned" animations in order to allow things like differently-shaped skeletons correctly interacting with things (varying body sizes correctly grabbing a cup), or using the same animations for differently-shaped skeletons without intersection issues? Or is there some other way of handling these things?

Edited December 19, 2017 by Dryspace

[Tchuck]

Anything non-human uses key-frames. Humans, it is a matter of reality vs style. If you want style, you'll have to go key-frame. If you want reality, and can afford mocap, it is far superior. If no, key-framing it will do.

 

And yes, we have things like inverse-kinematics that is used in rigs when animating so that you can properly position the limbs and whatnot. In runtime, we have variations of that to ensure better results. For example: when a character touches a wall when he's trying to move through a narrow area; placing the character's feet properly on the ground; moving the neck/eyes/shoulders to make the character look at something in a more natural way. All of those things are altered on runtime based on specific parameters. Since it would be humanly impossible to animate a foot touching all surfaces possible in a game.

 

Adjusting animations between differently-shaped skeletons is also done through something called retargeting. Basically, if your game pipeline is good, bones will be properly tagged on all your skeletons. Retargetting automatically looks for similar bones and attempts to map the existing skeleton to the new skeleton. Some cleanup is always necessary afterwards, though.

 

For intersection issues, depends on how important they are. Bones can also be constrained as to how much they can rotate and in which direction when you are making animations. So you already reduce potential intersection there. This can also sometimes be used on runtime if you have the systems for it.

I know EA had a very cool system for the first UFC, where the muscles were also dynamic and would match the incoming muscle up to a degree to avoid intersection, and push back if necessary.

[Gian_Yagami]

most of RPG games using separate body parts and attach it into same bone. this is image I found on gtainside for sample:

 

 

 

I found modeling clothes tutorial, if video games which have clothes change using this method (attach and fitting automatically on naked body) It will great and make 3D artist easier because they could share same clothes on different body mesh. And That was I thought clothes moving dynamically because it using cloth physics. And now I understand it is not, just additional bone like you tell.