Full body characters will not copy mouth shape from camera.

Ok, let's start with Smile and Surprised then. See if we can get them going. I am going to start with obvious things that I think are probably fine since it works with some puppets, but good to be sure! 😉

First thing, I assume you have the webcam on and can see your face up in the top right corner?

Next, if you look closely at your picture, after you have your face in the panel for a bit, do you see little red dots on your face? If so, that means it has recognized your face and is tracking it.

I think those two will be fine since it works on some puppets. So then open the puppet up in rig mode (I double click on puppet in project panel). Then select the root of the puppet and have a look for the Face behavior in the Properties panel to the right.

In the Face behavior, expand the "Handles" and "Replacements" sections. There should be a little ">" character to click on to expand. We want to make sure Handles has found "Head" and "Mouth" in particular. You can expand those sections to see the layer it the puppet they found and bound to. If they did not bind to something, that is a problem.

Next, expand the Replacements section and look for "Mouths". Under it I have "Mouth Group", "Neutral", "Smile", and "Surprised". I never worked out what Mouth Group was to be honest, but Smile and Surprised I wanted to make sure were bound to something. If not, that is a problem.

Finally, make a recording with the Face behavior armed. Make some expressions on the face to make a recording then stop. Try playing it back. Do expressions work then?

The above was to try and track down which step of the puppet has the problem. I very much doubt its full body vs head only - it is more likely to be something about the structure and tagging in the puppet. Hopefully the above will give some more evidence. Oh, a screenshot of the rigging panel of the puppet (expanding the mouth laters) could be useful too - so I can see the puppet hierarchy for one that is not working.

The camera controls two mouth positions - smile and surprised. The rest of the mouth positions are triggered by audio processing. It listens to the microphone doing its best guess of the sound and what letter is being said, then it picks the viseme based on that.

I am wondering if the "ee" mouth shape is actually a smile. Try doing a "reset pose" button near the camera. It might have picked a default face position that was not a good nueutral pose. So when you stop talking the web cam is guessing you are smiling. Reset post will hopefully help if that is the case.