Some ARKIT shape keys are designed to be combined together to create the desired expression.

• browInnerUp + browOuterUpLeft + browOuterUpRight
• cheekSquintLeft + eyeSquintLeft & cheekSquintRight + eyeSquintRight
• jawOpen + mouthClose
• mouthLowerDownLeft + mouthLowerDownRight
• mouthShrugLower + mouthShrugUpper
• mouthUpperUpLeft + mouthUpperUpRight

These expressions can be combined in a few ways, but generally speaking we can think of Shapekey A and Shapekey B creating a final Shapekey C. Or you could start from a Shapekey C, and split that one into a Shapekey A and Shapekey B. Where C is only used for production, and A and B are controlled by the motion capture. This latter option is more preferrable for some shapes, namely all the mouth shapes such as mouthLowerDownLeft + mouthLowerDownRight.

When splitting your Shapekey C into an A and B, each vertex shared between both A and B should have a value of C that adds up to 1.0 or 100%. So for example, that vertex might have a value of 0, meaning the resting pose, on A, and a value of 100% towards C on B. These splits are pretty complicated to balance. Generally, for a Left & Right combination expression, a vertex on shapeLeft would be closer to 100% the further it is to the left, and would be closer to 0% the further it is to the right. All these vertices should add up to 1.0 of C when A & B are both active, although this isnt always necessary.