I want to combine one animation for the body with other for the face. The problem is that despite creating an action in XSI with only keyframes for the corporal bones or facial bones, when I load it with the XSIXNARuntime, the animation seems to have one channel with at least one keyframe for each bone in the skeleton. So when I play the corporal animation and the facial animation with blend 1 for each one, the resulting animation is equal to the last I have play, instead of the combination of them.

Any idea how can I solve this?

Is your problem when you move to xna or is it in xsi itself?

In XSI I can mix the animations using the mixer and all is ok. My problem is when I move to  XNA. It seems like the content importer or content processor (I really don't know how them work) writes always one channel per bone with a keyframe at time 0.0 in spite of what you include in the keysets.

My simple solution to this problem has been to discard always the last keyframe of each channel, avoiding the unwanted effect of these frames that appears at each channel.