The actual animations are pretty good; they are just over the top. It's a common issue with motion capture because emotion in humans is very subtle, especially around the mouth area. If they reduced the range of motion on the mouth animations by about 60% they would get a much more believable result, but animators become fixated by the "sad must mean downturned mouth, happy must mean upturned mouth" trope, whereas in reality mouth movements are very small and implied much more by the nose, cheeks and eyes. Take an example of an animation you don't like and block out the mouth you will instantly feel like it's 3x better. A real world example of this is when poeple are trying hard to fake emotion they do the same thing - just watch vids of Amber Heard in court and you will instantly recognise it and get the exact same sense of disbelief and surreality...
