[Bf-committers] Collada importer/exporter kickout

Peter Amstutz tetron at interreality.org
Tue Jan 17 22:43:57 CET 2012


So the 2.61 COLLADA support isn't just bad, it's a regression from 2.59? 
  What changed?

Personally, I would be happy with reliable import of skeletons and 
rigged/skinned models.  My use case is to be able to move characters 
that were modeled/rigged/animated in 3DS Max into Blender, and not look 
back.

On 1/14/2012 3:28 PM, Juha Mäki-Kanto wrote:
> Hi,
>
> Did some digging into this ->  armature animation vs
> collada<http://www.pasteall.org/28205/text>
>
>> From testing and looking at the code Blender imports and uses animation
> curves directly which isn't good. It'll only match the original in special
> circumstances like if the rotation difference between bone's restpose vs
> parent's restpose is identity and you're only animating rotations. For
> Second Life there is actually a trick to set all restbones to be positive
> Y-axis (identity matrix) which apparently then loads correctly there.
>
> It seems that 2.59 did export/import for armatures correctly
> (#29082<http://projects.blender.org/tracker/?group_id=9&atid=498&func=detail&aid=29082>);
> tested it and the eulers in the file were interpreted to be quaternions
> (and the animation works) so the previous system was somewhat aware of this
> caveat.
>
> The gist to me is that collada's transformations may not match a program's
> internal representation and reinterpreting usually goes via collada data to
> full matrix, possibly fix the matrix and then decomposite to curves. It's a
> lossy conversion, key's should match but interpolation won't ->  you need to
> bake/key every frame when importing the curves.
>
>
>



More information about the Bf-committers mailing list