Actually it would be hilarous to find out how well the usual audio codecs compress an hour-long recording of a human repeating the same word over and over again :D
And comparing that to say an audiobook, to narrow the fields down to the frequency patterns and usual range of human speech.
In the end, searching for something akin to the gzip based approaches to text synthesis recently posted here :D
I mean, it seems obvious that the entropy of a recording of someone reciting the digits of Pi in english should be higher than the one of some poor person repeating "Zero!" over and over again...
I guess.
At least makes one chuckle, or have an attempt at it...
Or one could use it as bedtime reading: Zero, zero, zero... say it aloud and you're sound asleep.