Lifespans is electronic music where no actual voice sounds were used.
But if you listen to it at a fairly loud volume, or with headphones, you will seems to hear myriads of voices, singing little melodies that never repeat themselves.
It is based on a short sound sample made by frequency modulation. The sample contains audio below 20 Hz as well as above 20 kHz, both inaudible when it is played at the original speed.
But in Lifespans it is played at speeds between one-hundredth of that, and one hundred times that.
So both the sub-audio and the very high frequencies come into the audio range, and the sample gets a wide range of different ‘lifespans’.