Other than that, I also think you may want to hear a different polyphonic singer.
It's the same thing we do when we shape our mouths to make different vowel sounds, only much more precise. The physical thing that underlies what we perceive as "vowel quality" is the frequency of spectral peaks ("formant" is the linguistics terminology). When we shape our mouths (more generally our vocal tracts) to say "aaa" versus "iii" we are creating resonances at certain frequencies.
Admittedly when creating vowel sounds the bandwidth of the resonator is pretty wide, and when we hear vowels we can still perceive "a" quality vs "i" quality with even a wide bandwidth.
What Anna-Maria Hefele and other overtone singers do is much more precise. They can get a very narrow bandwidth by controlling their vocal tracts more precisely.