transfer shapekey give a crash

[bouich jules]

Hello

i am trying to transfer shapekey from an object in genesis 8 to the same object in genesis 9

‌

i get this crash:

‌

[Thomas Larsson]

Hm. I cannot reproduce this error. Which settings for the transfer shapekeys tool are you using? And what is the morphs you try to transfe?

[bouich jules]

i get even a longer crash:

‌

‌

i have put the file in attachment, try transfer shapekeys from top.001 to top

[bouich jules]

i couldn’t upload the file here it’s too heavy but here is it on mega:

‌

https://mega.nz/file/V1QziaRK#l_R774F898WKW4M1qMZGhcxDrvaKcsgg7Mrb0PiOrYU

‌

[Thomas Larsson]

Since the only mesh with shapekeys is Top.001, I presume that you want to transfer shapekeys from that mesh to the other top. That indeed causes a numerical error when solving a large equation system, presumably due to limited floating point precision. The plugin now catches that error and suggests that you should use the Legacy transfer method instead. It is slower but more stable than the Nearest Face method, at least in this case.

Another problem is that the two tops have different armatures. The transfer tool only works properly if the two meshes are at the same location. Therefore object transforms are not allowed, but you can circumvent that check by parenting the meshes to different objects. Now this results in a warning. You might have good reasons to transfer between meshes with different parents, but in that case the parents should not have object transformations.

If you enable the Transfer Drivers option, the shapekeys are driven by the other armature (Genesis 8.1 Female). This is probably not what you want, so it is better to disable this option and use Advanced Setup > Morphs > Add Shapekey Drivers afterwards.

[Alessandro Padovani]

@Thomas, I found this on google, it explains the error raises when you try to invert a singular matrix that can’t be done by definition.

https://www.statology.org/python-numpy-linalg-singular-matrix/

[Thomas Larsson]

Yes, but I think the matrix becomes singular due to rounding errors. If we could work with infinite precision it would probably work. Anyway, the error is caught and there is an alternative method which is more stable.

[Alessandro Padovani]

If it can be useful here is an article where it explains that you can use pseudo inverse to avoid the error.

https://stackoverflow.com/questions/52289379/invert-singular-matrix-on-python

https://numpy.org/doc/stable/reference/generated/numpy.linalg.pinv.html

[bouich jules]

thank you!