Spectral slices of overtone singing, animated

As part of my on-going exploration of the many ways in which F0 is not pitch and pitch is not F0, I did a little demo/experiment with a sample of Anna-Maria Hefele's "Polyphonic Overtone Singing" video:

Here's an 8.85-second audio clip from that video, starting at 43.45 seconds:

Your browser does not support the audio element.

Here's a spectrogram of that segment:

(The blue line at the top is Praat's attempt at pitch tracking, which follows the fundamental (at about 275 Hz) rather than the variable-strength harmonics which we hear as a time-varying melody.)

And here's the fun part, an animation of sequential spectral slices from the same clip. I've slowed the performance by about a factor of three — the spectral slices were calculated 100 times per second, in overlapping 15-millisecond windows (code available on request), whereas the animation runs at 30 frames per second. I extended the audio to match the duration of the video, without changing the pitch, using sox's tempo feature (i.e. "sox Hefele1X1a.wav Hefele1X1b.wav tempo 0.3051724 50"):

Your browser does not support the HTML5 video.

More later on what this means (for those who don't get it already), and applications of the same kind of visualization to "creak", "throat singing", "voice quality", etc. etc.