Sort of interesting. But in the demo, contrary to what the narrator says, it's not a "b" that is being pronounced, but rather a strongly articulated unaspirated "p". This is not a sound that occurs initially in English (it does occur after initial "s" in words like "spell"), which may be important to the illusion. You hear a sound that is ambiguous -- it's a little like several consonants pronounced with the lips, but not quite like any English sound, and so you need the visual cue to identify which English sound is intended. I wonder whether a Hindi speaker would get the illusion, since Hindi has both aspirated and unaspirated "p" phonemes.