Just to update with my own findings even if I'm talking to myself here ;)
I have a decent test motion capture file that's good enough to use in order to try and get the end-to-end chain done. I also now know the exact workflow to get an SMD file into Source:
Decompile an MDL file with CannonFodder's decompiler tool Add your SMD file to the same folder as the rest of them Edit the qc file to take into account your new SMD file Recompile the qc file using StudioMDL.exe
It's not as easy as that though, as the exported SMD file from IPISoft isn't compatible with Source as it currently stands... To get it into the right shape I have to: 1) Edit the SMD file in Notepad to change bone names and tidy up one missing and two surplus bones. More detail on this once I've proven the process 2) Change the bone lengths. While bvhacker replays the file perfectly, the lengths are about 30 times too short for Source, so your model ends up a scrunched up ball. Again, more info on this as and when I find it out. 3) The rotation values are probably screwed as well, because the result looks like a teleporter accident. I can't get much further with troubleshooting this though, because... 4) ...compiling my qc files fail miserably anyway.
The thing with 4) is, I can decompile a .mdl and get the qc and all associated SMD files out of it, and they look to my eye intact and functional. But if I then recompile that qc file WITH NO CHANGES (i.e. I haven't added an SMD or edited the qc file), the resulting .MDL files is a) about three times the size of the original, b) can't be decompiled again, and c) is completely broken - every gesture is damaged with legs pointing the wrong way, extra gestures appear that shouldn't etc.. Not very useful.
So now I'm trying to work out what's kerfucked on the Source side that stops me from even simply decompiling and recompiling a model successfully. An uphill struggle :(
-Ian
|