|
Totally agreed...
I was just spreading the word for the rigger here (who up until now is
quite happy with what he sees...)
(still: what he noticed was this strange behaviour on scaling which
doesnt seem to occur in the other implementations...)
Bradley Gabe schrieb:
Dual Quaternion skinning offers a different mathematical approach to
solving for deformation, but it alone by no means is *the* solution.
It's the proverbial procrustean bed, some things improve but at the
expense of others. While volume is better maintained at joint centers
and candy-wrapper twists go away, the resulting deforms still don't
look natural, and in some cases look worse than what you get with the
traditional envelope algorithm.
The main issue for us as riggers has always been trying to figure out
clever ways to isolate our rotational axes per joint in order to
distribute different types of deforms per joint motion. Example, an
elbow rig. Assuming a standard XSI IK chain, we bend the elbow by
rotating the second arm bone in the z axis, and twist the forearm by
rotating it in the x axis. The elbow bend is best if the effect is
localized close to the joint center. The forearm twist, we would rather
distribute down the length of the forearm towards the wrist. A dual
quaternion solution doesn't isolate those two deforms, we will still
have to do that via rigging. And while the dual quaternion solution
will keep our forearms from losing volume as they twist, it's going to
be debatable as to whether they look better on the elbow bend.
Here's where the future actually looks bright for those of us without
cMuscles built into our software. The more ideal solution for the
deformation problem does not lie in an all or nothing algorithm, and
possibly not in someone else's pre-canned muscle system. I suspect, the
more ideal solution will be found in a conditional deformation system.
One where we get to choose from many different types of deforms and
have them activate, blend, and normalize based on a set of kinematic
rules... When the forearm rotates in the z-axis, apply an 80-20 blend
of dual quaternion to traditional envelope. When the elbow rotates in
the z-axis, apply a 20-80 blend, along with a localized push
operation... Basically, we're taking all the jiggering away from the
externals of rigging and moving them to the internals of the
deformation system where they can be out of the way of the animators,
and nicely multi-threaded. :-)
On Dec 18, 2007 9:24 AM, Philipp Oeser < philipp.oeser(at)liga01.de>
wrote:
Hi,
it really is an improvement over the "regular" envelope, thanx
again...slowdown can be noticable but what the heck...
One thing though:
Scaling seems to be quite different than the "regular"envelope.
Ladislav
Kavan talks about solving this limitation in his realtime approach by
having extra vertexshader instructions.
Is there something like this in the current implementation?
In any case: thanx!
Philipp
Fernando Navarro schrieb:
> Hi all:
> I have released a new envelope operator that implements dual
> quaternion interpolation. It has been designed to replace the
existing
> envelope operator using the same inputs (deformers, weights, ...).
It
> can be used with NURBS surfaces and curves, polygonal surfaces and
> lattices.
> As part of the package I have written a brief explanation that
will
> help you understand the differences between the traditional and
this
> method.
>
> You can download it at http://www.highend3d.com/f/4977.html
>
> It's a working release but can be improved. If you have and
> suggestion or find a bug, just drop me a line...
>
> Hope you enjoy it!
> Fernando
> ---
> Unsubscribe? Mail Majordomo(at)Softimage.COM
with the following text in body:
> unsubscribe xsi
>
>
--
LIGA_01 COMPUTERFILM GmbH
Philipp Oeser
MUNICH
Landwehrstrasse 60-62
80336 Munich, Germany
fon: +49 - (0)89 - 51 555 8-88
fax: +49 - (0)89 - 51 555 8-77
www.liga01.com
Diese Nachricht enthält vertrauliche und/oder rechtlich geschützte
Informationen. Wenn Sie nicht der richtige Adressat sind oder diese
Nachricht irrtümlich erhalten haben, informieren Sie bitte den Absender
und löschen diese Nachricht. Das unerlaubte Kopieren und die unbefugte
Weiterleitung dieser Nachricht sind nicht gestattet.
This message may contain confidential and/or privileged information. If
you are not the intended recipient or if you have received this message
in error, please notify the sender and delete this message. Any
unauthorised copying, disclosure or distribution of this message is
strictly forbidden.
---
--
LIGA_01 COMPUTERFILM GmbH
Philipp Oeser
MUNICH
Landwehrstrasse 60-62
80336 Munich, Germany
fon: +49 - (0)89 - 51 555 8-88
fax: +49 - (0)89 - 51 555 8-77
www.liga01.com
Diese Nachricht enthält vertrauliche und/oder rechtlich geschützte Informationen. Wenn Sie nicht der richtige Adressat sind oder diese Nachricht irrtümlich erhalten haben, informieren Sie bitte den Absender und löschen diese Nachricht. Das unerlaubte Kopieren und die unbefugte Weiterleitung dieser Nachricht sind nicht gestattet.
This message may contain confidential and/or privileged information. If you are not the intended recipient or if you have received this message in error, please notify the sender and delete this message. Any unauthorised copying, disclosure or distribution of this message is strictly forbidden.
---
Unsubscribe? Mail Majordomo(at)Softimage.COM with the following text in body:
unsubscribe xsi
|
|
