Credits: Matthias Zauner, antokhio, microdee
Latest version requires the DirectX11 pack.
This is Instanced Skinning in dx11 with thanks to microdee for the shader.
The module is still keeping track of all your assets if you asign it to a folder. This is nice for the workflow.
In this version I step away from any attempts to make it kinect-ready or provide a rigging or animation framework within vvvv.
This is only for animating multiple instance of one colladda file.
The Module was created before vux published fbx and better collada support in the assimp nodes, so basically it is outdated already.
However it can be useful if you want to have a definite workflow or if you want to reuse your old collada meshes, since it uses the old dx9 collada node.
Exporting Collada files in a way so it works with this module was hard to find out since many export chains do not work.
What workes 100% is:
Export from Cinema3d as FBX
Open in 3dsMax 2010 or 2011 and export again as .DAE (Open Collada)
The Open Collada Plugin can be found here https://github.com/KhronosGroup/OpenCOLLADA/wiki/OpenCOLLADA-Tools
This always workedd for us.
I have to point out that in the meantime the Assimp Nodes provide better functionality and I saw people using directly FBX or using .DAE files exported from native blender exporter. I point to microdees emeshes to see an example of this.
NOTE for the old dx9 versions that are extreamly outdated: You have to download the test models separately and extract the folder MODELS in the contribution folder.
anonymous user login
Very nice! I will try with the kinect.
In CustomSkeleton (AssetsDB).v4p I have Strider node which is red.
sonostrano: I had same problem, load the strider node from inside its folder and then red color disappears and all working well. Let me know if you still get problem.
TEKCOR: Thankx a lot for this contribution man, it looks dope! I am playing around with it now and will report any problem or bugs. I am sure I will figure it out but if not could you just briefly explain how to link the kinect skeleton with this animatedCharacter skeleton? would that be with bindskeleton? I have been trying to do basic mocap using VVVV + kinect for about 2 weeks now and it's driving me insane. As Antokhio has been posting on all related threads it is not an easy thing to do, and I also seem to come across bone orientation problem, and I am hoping your module somehow makes it work...
Thankx a lot again for all your work, it is much appreciated!
Hey jo evvvvil yesterday I worked on the kinect thing with newemka. Cool thing is that the OpenNI nodes give a rotation and we should be able to map it to the right joints. Unfortunatly OpenNI nodes doesnt work for me somehow. Have to figure it out. But in theorie the only thing you have to do is to download the new version, hook up the kinect, check out inside the Skeleton (OpenNI) module which of the joints doesnt go into the orientation spread (because they are the last joints and dont have a rotation) and then you have to go very deep inside the KinectJointMapping module and make a new filter spread for the sift. So that the KinectJointOrientation Spread is going into the right order into the Rotate node. And of course it has to have the same spreadcount then the filter spread of the sift.
Cause the problem right now is that the orientation spread is smaller then the spread with all joints.
newemka fortunately already selected the right order of joints as it gets output by the Kinect (Microsoft) Node.
Also add me in skype. I have time again this evening at 7 or something.
e: and dont forget to copy the MODELS from the old folder or download it separately from here. I made it outside of the versioning, because it is 20 MB.
Maybe it's because the knees don't need to rotate in every directions, like the elbows
Word man, thankx for your reply. I will add you on skype around 7pm tonight and I be on this too. I am trying to make sense of everything in your module, forgive me only been using VVVV for 2 years... Your module is great though, don't get me wrong, SO COOL. BUT a few things: Don't I have to somehow redesign the Strider custom skeleton thingy (to fit with openNI skeleton)... I must say it starts making sense when I go into your JointMapper node, I am guessing that is what you mean by kinectJointMapping module, but then hasn't your skeleton got 4 extra bones? shall I kill those as the list of openNI bones wouldn't include wrist or ankle... I get what you are saying about joint orientation on kinect openNI not working on end joints, that's true and very valid point. But aren't all bones returning an orientation regardless of whether it is end joint or not? From the skeleton openNI node I seem to get an orientation for all joints out of it (X axis orientation, Y orientation, Z axis orientation)... Let's talk tonight man, I am keen on finding a working workflow to do motion capture using the kinect in VVVV on a collada model. And I mean a REAL workflow which works from A to Z, from how to rig the model in Max/Maya etc to how to export it in Collada (which setting) and the actual VVVV patch linking correctly the kinect OpenNI Skeleton to the collada model's skeleton. We all want to do this, I wanna do a video tutorial explain the whole process as soon as we have a found a fully working workflow... PEACE MAN!
yes cool. building a motion database from the kinect tracking is the way to go.
but to clear some confusion.
Ignore the BindSkeleton and the CustomSkeleton and the Strider for now.
This is a developement tree I am not finished with:
In the custom skeleton there is a patch switcher that can load skeletons that were created in vvvv. In this case it loads a patched skeleton for a Half Life 2 Strider model that I found. It doesn't work so far, but this is another topic.
for the Kinect mapping we look in version 0.3 I uploaded just a few minutes ago in the AnimatedCharacter->SkeletonAnimation->Animation->KinectJointMapping.
okey sorry for everyone who downloaded it and probably got confused.
The joint namings were not assigned correctly to the incoming rotations from OpenNI. After using my FindIndexAssociaton which I created for these kind of database operations, it works now.
But still: this module doesn't solve the general orientation problem. It just makes it easy to use different models with different animation sources.
any development on this front? I ve tried different way to get the Kinect to animate a character but to no avail...
nope nothing here. Sorry. Still some brain and time has to be invested to change the transform matrix in the correct way.
I tried different ways but not able to animate yet.
I suggest it may have a possibility to get from Mocap the flux of skeleton ?
Or may Iclone able to send this flux or directly the result with caracter to vvv via osc ??
Like a bottle in the sea...