Autodesk FXB export issues

Hi,
I’ve recently been investigating the FBX export script, trying to export the bird from yofrankie into XNA using the skinned model sample from XNA creators club.

As is widely known, the current bundled export script does not work at all whilst the one from tripleb games works a little better (The birds exports okay but the sheep is still wrong).

My investigation highlights 2 differences:

  1. The blender export script sets the unit scale to 100 whist tripleb sets it to 1. The problem here is with the XNA content pipleline in the sample - it does not bake the overall transformation matrix. It’s a trivial fix in the XNA sample. Still, blender should not hard code 100 here - if anything it should be the value set by the scale slider in the GUI or 1.

  2. The matrices that blender writes into the BindPose element of the FBX are wrong. The tripleb script uses a different value here (see my comment at the end of the other thread here: http://blenderartists.org/forum/showpost.php?p=1415476&postcount=13).
    This gives a better value, but it is still wrong.
    The confusing thing seems to be that the FBX viewer for quicktime ignores the BindPose - so it looks okay in there. XNA uses it though.
    Luckily, at least for the XNA pipeline, this seems to be optional. Deleting the BindPose element entirely seems to fix everything and the FBX still shows okay in the quicktime viewer.

Ideasman … any comments?

Ian

the FBX exporter works fantastically for maya and Lightwave… (not a max user so can’t talk there)

I guess the unit scale is for good transfer to other editing software… It’s consistent with OBJ and LWO imports…and comes through as 1 blender unit == 1meter which is how most people use it…

So either separate exporters or an export dialogue to set this I guess…

ah - the one unit per metre makes sense.
I suspect that maya ignores the BindPose and just recreates it from the skeleton definition. Certainly, the quicktime viewer (from autodesk) seems to ignore it - you can edit the values to whatever you like and it makes no difference. The XNA importer does use it though if it is there.
I think the bind pose needs to have the inverse matrices in it and blender is writing the actual transform matrices … I’m not sure on that though and I don’t have max, maya or the FBX SDK to investigate.