MetaHumans - Expression as a combined blend shape used for phonemes


is there any way to create a blend shape based on a created expression using the control rig?

We need to create a set of specific expressions for speech, as shown in the example here (expression we would theoretically use for /UE:/ (As in You))

The issue we have is that to link the specific blend shapes that exist would be a major hassle and we are trying to figure out if we could create a combined blend shape that we could easily reference in speech recognition. If anybody is able to point us in the right direction, that would be incredible.

We tried to create a combined blend target in Autodesk Maya, but it seems the blend shapes are bound to the rig itself and we could not figure out how to make it work.

Hopefully I am not asking a silly question with a really simple answer, if so, please, school me :slight_smile:

Best regards,
Daniel H.

Have you had a look at tge arkit map pose asset used in the facial anim bp that comes with the metahunan project? I think you should be able to use a pose asset for this.

I mean … the blend shapes exist in the skeletal mesh, not here, or am I wrong? :open_mouth:

If you could sort of guide me into how you mean that exactly? I can create the blend shape in the imported source asset in Maya, but I am having an issue with importing it back into UE. If I can import it into this pose asset for anim without it breaking bone connections etc, that would be cool, but I doubt that :frowning:

As far as I understood your initial requirements, you model a phoneme pose in the control rig… so you are adjusting various sliders until you get the desired expression.

Then I gues you want to animate the character to make these phonemes (make her talking).

The sliders in the control rig correspond to curve values that drive blendshapes. So you could create a pose asset with poses for all your phonemes, which in turn set the correct values for several of the blendshapes/morph targets. It’s not really creating a new blendshape stricly speaking, more like grouping sets of values, so you can control the single set.

For example your phoneme UE consist of the following values
CTRL_morphA 0.2223
CTRL_morphB 0.723
CTRL_morphC 0.21

You could create a pose “UE” for this. If your set the curve value for UE to 1, the character’s face make the UE expression.

1 Like