So I retackled the bone issue (wow, how long has it been, over a year?) and recorded some progress. Apparently, of those 39 bone bytes (counting from 0) the bytes 7-14 are quaternion values and the next 12 are xyz values (similar to vertices). I've updated the first post.
It's still missing something though. The resulting skeleton fits some models perfectly:
but for some models, some parts aren't rotated properly:
Second one seems like the entire skeleton is tilted but actually, the bones related to the clipboard on her hand are fine. That's why I said "some parts".
Well, if there's anyone interested, feel free to check and experiment with this new info.
I also had a look at those newer files, they seem to be the same format in general but some have slightly different PMM chunks which cause problems for existing scripts. I could check further but I think it would be better if we solved these bone issues first.
It's still missing something though. The resulting skeleton fits some models perfectly:
but for some models, some parts aren't rotated properly:
Second one seems like the entire skeleton is tilted but actually, the bones related to the clipboard on her hand are fine. That's why I said "some parts".
Well, if there's anyone interested, feel free to check and experiment with this new info.
I also had a look at those newer files, they seem to be the same format in general but some have slightly different PMM chunks which cause problems for existing scripts. I could check further but I think it would be better if we solved these bone issues first.