Re: OT: Autodesk purchases cMuscle

["Bradley Gabe" <withanar(at)gmail.com>]

In theory it is a good idea, in practice, it has a ways to go. It has the same main issue we have using quaternions for rotation animation in that it's mathematically sound, but it is very difficult to create a visual representation for manipulation that makes intuitive sense to artists. So we control everything in Euler space, but allow quaternions to do the work behind the scenes. Now, apply that to envelope weighting

 

On 10/25/07, Ahmidou Lyazidi <ahmidou.xsi(at)gmail.com> wrote:

About the fact that we need a better enveloppe operator, have you guys heard about dual quaternion skinning? It seems to me a really good solution:

http://isg.cs.tcd.ie/projects/DualQuaternions/

2007/10/25, Bradley Gabe <withanar(at)gmail.com>:

I have yet to see an off the shelf solution for muscles in any 3d software that I would want to use on a day to day basis in production. Granted, if I were working on a Discovery Channel special about the anatomy of motion, I would definitely want a copy of something akin to cMuscle, but only to generate the actual muscles and not any flesh deformation outside of it.

 

<soapbox>

Until 3D envelope deformations are better solved to maintain volume around the problem areas (shoulders, hips, candy wrapper twists, etc) by default, any secondary system for muscles and fat is going to fail in a typical production schedule. For these systems to work efficiently, you need a baseline body deform, and all the secondary stuff, push, slide, jiggle, is applied as a delta on top of that. If your baseline is too far off, and most of the time it is, the secondary stuff ends up stuck in a garbage in, garbage out situation. That is why you'll see many examples of a bicep flex, which is easy to control, but few examples of arms being raised above the head and waving at the sky.

 

Instead, we are stuck in a paradigm where the baseline volume must be corrected *before* any secondary muscle/flesh sim is applied. And then quite often another corrective pass *after* the muscle/flesh sim is applied. Smaller scale productions simply don't have the talent or budget to handle this level of work. And ultimately, many of the same kinds of secondary looks can be reasonably achieved in less time with simple spring jiggle and shape animation.

</soapbox>

What should be interesting about Moondust for us is that it should provide a context where we can hopefully (and finally) solve the base volume deformation issues. We should be able to create a "smart envelope" that behaves one way when our character does a simple bicep flex, and a different way when the forearm twists, and yet another way when the arm is raised above the head. Many of us already do this kind of conditional deforming with envelope nulls that move or scale according to constraints or expressions. But access to the deformer definition from the inside, and driving *that* with conditional relationships is going to raise our abilities to a whole new level.

 

-Brad

 

On 10/25/07, André Adam <a_adam(at)49games.de> wrote:

Whoa, no, I would not even think of doing that with the whole body skin;
only with selected vertex clusters I consider safe area, blended with
weight maps. On top of my head Shrinkwrap casting along vertex normal,
with the bidirectional flag set. Again, not an approach for flesh
simulation; muscles only.

   -André


Frank Lenhard wrote:
> where you able to use shrinkwrap to drive a whole body skin?
> my tests always lead to intersection and flipping problems, using a
> nurbs "muscle" conencted to the bones driving the higrez mesh of a
> char using shrinkwrap.
> at some point, rather sooner then later, the muscle intersected the
> skin and it flipped.
>
>
> ciao
> franky
>
>
> Thursday, October 25, 2007, 11:40:33 AM, you wrote:
>
> AA> Not talking Syflex here, the post Wayne quoted from aims at deformable
> AA> objects (shapes, push, etc...) linked to the primary skeleton with the
> AA> envelope being shrinkwrap-projected onto them as a secondary
> AA> deformation. This can be set up without any plugins using the software
> AA> in front of you.
>
> AA> Since this approach is not simulation-based, one can argue if it should
> AA> be called a "muscle simulation", though I have to say that I'd go that
> AA> route anytime in favour of a simulation-based approach because it
> AA> creates stable and reproducable results.
>
> AA> If you need your character's envelope to behave jelly-like, now, that's
> AA> a different story, but not related to Wayne's post quoting from the
> AA> moondust demo.
>
> AA>     -André
>
>
> AA> Matt Lowery wrote:
>
>>> Again this comes back to the muscle system that xsi currently has but no
>>> knows how to use, or where to go to learn how to use... Syflex! Skin and
>>> flesh dynamics... built right into xsi. But to be honest I'd be massively
>>> surprised if anyone on the list has ever used it in production because it is
>>> just so badly documented.
>>>
>>> -----Original Message-----
>>> From: André Adam [mailto: a_adam(at)49games.de]
>>> Sent: 25 October 2007 08:28
>>> To: XSI(at)Softimage.COM
>>> Subject: Re: OT: Autodesk purchases cMuscle
>>>
>>>
>>> No need for new bells and whistles then, you can already do exactly that
>>> with the current toolset.
>>>
>>>     -André
>>>
>>>
>>> Oz Adi wrote:
>>>
>>>
>>>> yes, that's the one :)
>>>>
>>>>
>>>> Wayne Williams wrote:
>>>>
>>>>
>>>>> I think you are talking about adam's post.
>>>>>
>>>>> "6) muscle sim.. simple arm flexing example with underlying muscle
>>>>> object driven by an f-curve node that could have its profile updated
>>>>> on the fly to change the muscle shape.  controls for changing muscle
>>>>> height, thickness, compresison, min, max etc.. uses a raycasting /
>>>>> shrinwrap type node to have the high res mesh follow the underlying
>>>>> muscle.."
>>>>>
>>>>> I guess we'll see what is in store for XSI in the coming months.
>>>>> -wayne
>>>>>
>>>>>
>>>>> ----- Original Message ----- From: "Oz Adi" < oz(at)broadcast.co.il>
>>>>> To: < XSI(at)Softimage.COM>
>>>>> Sent: Wednesday, October 24, 2007 1:08 PM
>>>>> Subject: Re: OT: Autodesk purchases cMuscle
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>> maybe :)
>>>>>>
>>>>>> but as far as I can remember, there was a thread here, that someone
>>>>>> wrote from the top of his mind, what he remembered from
>>>>>> the moondust presentation at the user group meeting... and I am
>>>>>> almost sure he wrote something about a muscle system...
>>>>>>
>>>>>>
>>>>>> peter boeykens wrote:
>>>>>>
>>>>>>
>>>>>>> let me have a guess... they showed CAT for 3dsMax?
>>>>>>>
>>>>>>>
>>>>>>> ----- Original Message ----- From: "Oz Adi" < oz(at)broadcast.co.il>
>>>>>>> To: < XSI(at)Softimage.COM>
>>>>>>> Sent: Wednesday, October 24, 2007 6:14 PM
>>>>>>> Subject: Re: OT: Autodesk purchases cMuscle
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>> I remember reading a couple of months ago, a report from siggraph,
>>>>>>>> that soft showed some kind of muscle system
>>>>>>>> that is related or not related to moondust ?
>>>>>>>>
>>>>>>>>
>>>>>>>> Wayne Williams wrote:
>>>>>>>>
>>>>>>>>
>>>>>>>>> http://www.cometdigital.com/cMuscleSystem_notice.php
>>>>>>>>>  well, i guess any hopes of that being ported to xsi are
>>>>>>>>> squashed. Is this even replicatable in XSI with its current toolset?
>>>>>>>>>

---
Unsubscribe? Mail Majordomo(at)Softimage.COM with the following text in body:
unsubscribe xsi