Wy does blendshape conversion produce different results than original model

[andrew199]

Got another question for you wonderful folks on the A2F team :)

I’m doing some testing and I’ve noticed that blendshape conversion results in slight differences when compared to the original model. I’ve attached a screenshot of the male_bs_46.usd blendshape model with default settings as an example of it being different.

Is this because the correspondence in skin mesh fitting is not being done accurately enough?

So theoretically, if we were to perfectly align the correspondences on the skin mesh fitting, would it look exactly the same?

Also, if I wanted to adjust things after blendshape conversion, what are some ways for me to finetune it so that I can make it match better?

Thanks so much!

[yseol]

Hi @andrew199,
Thanks for trying this. The slight difference you saw is normal. Blendshape conversion is a lossy conversion, 46 or 52 poses can recreate the source motion to some extent but not perfect.

There are blendshape solve options sliders where you can adjust to make the resulting expression less strong, smoother, and symmetrical. You can use them by checking this video.

The main purpose of blendshape conversion is generating transferrable weight values that can be used on other characters. If you want the accurate face motion in the source face, the geometry cache export can be useful.