Hi, I had a mess about with getting the new BVH importer working but it was causing bugs with makeEditable, since I had started my own bone importer, I decieded to try and finish it. didnt finish it so wondering oif anyone else would like to have a go. all the problems start at line 320 so you can mostly focus on getting the empty location/rotation onto the bone.
Heres as far as I got, Rotation ‘Almost! works’
http://members.iinet.net.au/~cpbarton/bvh_import.py
Hotd the line! bvh+bones working now… Ill keep ya posted!
Cool! If you need any .BVH files to test, or want someone to test the script, let me know.
Very cool! Will it work with both types of .bvh files (Poser and BioVision)?
I also have some .bvh files created by MotView 0.8 that I can’t get Blender to read. Let me know if you could use more specific info about the file structure.
Was the “junk” error with the Joeg code due to makeEditable, then? Just curious…
yep it was todo with nmake editable, my script has no troubles making the armature (uses a funky generic dict2armature function I wrote) - but!!!
I have bow tried like 4 ways to get the rotations in and all have failed! Armature matricies is some kind of black magick!
Heres a cheating method, track the armature to the empties…
http://members.iinet.net.au/%7Ecpbarton/bvh_import.blend
Can anyone see if they can remove the empties but keep the animation, either by baking the action or using python… I had a go but I dont know how to yse the NLA/Actions very well… At least the armature now follows the BVH properly.
Can anyone see if they can remove the empties but keep the animation, either by baking the action or using python…
I think thats a question we’ve wanted answered since years ago. I am not sure if these motion capture files record the roll of any point, otherwise you might more or less copy the rotation of a bone from the empties or the direction from the difference in their position. I’m sure others here know more about the format. Let us know if you are successfull.
I’ve been fighting with rotations in pose import. The best I’ve been able to manage is swapping euler values to switch rotation orders. I’m sure you math guys can do better than that. 
But I’ve noticed that the chief problem is bone roll. In converting Poser .bvh motion or quaternion export poses, I’ve had to adjust the data by the joint order in Poser, pre-export, and again by the roll settings in Blender before it begins to look good. A point of importance has turned out to be the primary axis along which a bone points.
Your script worked well, aside from having the same rotation problems I’ve seen so much before. 
I hope you can crack this!