Page 3 of 3

Scripted Animation demo video

Posted: Sun Aug 17, 2008 6:53 pm
by westine
this demo uses the following script

Code: Select all

States: 7
Transitions: 6
larger pic
YouTube video

Posted: Mon Aug 18, 2008 4:42 pm
by KungFooMasta
I bet it looks really good normally, but the video is very choppy, I couldn't really tell what was going on. :(

Re: Scripted Animation demo video

Posted: Thu Aug 21, 2008 9:06 pm
by Tomat1979
westine wrote:this demo uses the following script
I think nobody undestend that's happen.

It should better to split screen on 4 part and in 1,2 parts show - chracter animation FROM (1 part of screen) and TO (2 part of screen) transation will happen. And in 3,4 parts show result of simple blend (3 part of screen) and result of blend using motion grapth (4 part).
Then people might see difference of simple blend and blend using motion graph.

Posted: Thu Aug 21, 2008 11:31 pm
by Wolfmanfx
Yeah westine you have to sell your work a little bit better :)

Posted: Fri Aug 22, 2008 9:36 am
by westine
Tomat1979's suggestion is totally accepted, there will surely be improvement, in which the "nobody" will be solved.
But now there the most bothering problem is character invisible phenomenon.

I debug it. In OctreeSceneManager::init in OgreOctreeSceneManager.cpp
I enabled setDisplaySceneNodes( true );
setShowBoxes( true );
to show the octree boxes and all scenenodes.

Image normal situation

Image[/url] character disappears and one three colored layer of floor disappears too

Posted: Fri Aug 22, 2008 11:45 am
by tuan kuranes
Exporter usually take care of including all animation in bbox.

If motion graph generates new animations, you may have to expand bounding box of entity yourself, as bbox may be only large enough to contain all backed animation.

Posted: Tue Sep 02, 2008 7:57 pm
by alexjc

Just caught up with this project. Looks interesting.

For future reference, you typically cannot bind arbitrary motions on existing skeletons without retargeting the whole animation (see Mike Gleicher's work). That's the sad thing about all the animation in CMU's mo-cap library, it's all with different skeletons so the price of entry is a motion retargetting tool if you want to combine the animations together.


Posted: Tue Sep 02, 2008 9:27 pm
by alexjc
Also, where does the BHV exporter live? I looked for it but I couldn't find it in the repository or your RAR bundle...


Posted: Wed Sep 03, 2008 8:58 am
by tuan kuranes
@alexjc: indeed motion targeting for those without mocap hardware is needed, but we had to restrcit gsoc subject to realistic goals.
Once we have set current code to main Ogre, that would sure be an important step (even demo would become sexier in this case.)

About BVH, it's in OgreXMLConverter exe tool, and in the

Code: Select all

OgreXMLConverter.exe -addanimation rush.bvh jaiqua.skeleton.xml rush.skeleton.xml 
direct link to svn bvh code

Any further direction guidelines, code review, or comments on the new code from someone in the field like you would be greatly appreciated.

Perhaps automatic rigging and other skeleton extraction like 1 & 2 would be of some help, ideally with some 'loddable' skeleton would be the perfect complement ?

Posted: Wed Sep 03, 2008 9:06 am
by alexjc

I understand totally about the restrictions on SoC. I was pointing it out to prevent you spending too much time trying to get the CMU mocap BVH working on Jaiqua!

Thanks for the link, I'll take a look :-)


Posted: Mon Sep 08, 2008 11:54 am
by alexjc

I will take a look at the whole code and let you know. I foresaw a few necessary changes in the animation states code to support a real blend-tree, but I've not checked out Westine's code yet.

Also, what's the full pipeline you guys use for creating those dummy skeletons before adding the animations? I'd like to try it out with a few different animations!


Posted: Tue Sep 09, 2008 1:18 pm
by tuan kuranes
I think Westine used Blender to create that dummy skeleton.
Westine, Perhaps you can provide blender source ?

thanks and future work

Posted: Thu Sep 11, 2008 10:36 am
by westine
@alexjc: Thanks for your attention. I am very glad to know a guy from, I have subscribed to this site almost for two years, as I am always excited with AI stuff even beyond games.
To get this gsoc project totally practicable, thanks to Ilya Baran's code.I have already looked into autorig and tried to bind cmu's skeleton.
The skeleton embeding and animation format conversion seems to cost at lease one month work.
Motion graph improvement will be added after my current teacher assigned work has been accomplished.
Source code is sent and thanks.