Yeah -- we have the participants use chinrests as well, which reduces head motion artifacts for typing but less so for speaking (because they have to move their heads for that of course). so a lot of the data is with them keeping their heads quite still, although the model is becoming much more robust to this over time.