Well let’s assume there’s a functionally infinite number of sounds and a functionally infinite variety of songs and voices. The set of songs and voices is a smaller infinite set than the set of all sounds. That would make the odds per generation effectively n/∞. So essentially zero.