I recorded myself singing the choral pronunciations of a, e, i, o, and u (same as Spanish, basically). Then I selected the most interesting bits, which tended to be when I was running out of breath and the pitch was wobbling, or when there were weird shifts in the overtones.
I made an interesting motive by mapping the samples to buttons on my Push and improvising a bit. Except - uh oh! - I had 9 samples and the Push grid only goes as far as 8. What to do? Well, I put samples 1-8 on channel/track 1, sample #9 looped around to to be the first sample on channel/track 2, and then I put samples 1-7 on channel/track 2 again. Then I played my cool little motive again, this timing using two fingers to trigger 2 samples at a time instead of 1. This layering made the track much more interesting - each sound repeats often but are seldom paired with the same sample, so it’s always slightly different.
I used the physical gesture of the first motive to generate the rest of the piece, especially the end. In Ableton terms, I switched between “scenes” 1 and 3 a few times, then let 4 last a bit longer. I generated the end of the piece treating that as a sequence (1,3,1,3,4 then 2,4,2,4,5 then 3,5,3,5,6 etc.)
True to the assignment, I did very little vocal processing. I did add a limiter to bring down the volume jump from my attack on the “a” vowel (I forgot to put it on the other track, so you can really hear the difference at 1:04 vs. 1:14, if you’re so inclined). I also added a touch of reverb.