[xwintermutex]

I guess it is because of the use of amplitude modulation [1], where the envelope ("strength") of the carrier (that is some higher frequency signal) follows the signal that is modulated onto it (speech, in this video). With say FM it wouldn't work, I presume.

[1]: https://en.wikipedia.org/wiki/Amplitude_modulation

[jimmaswell]

If you listen to an FM signal interpreted as AM you can often still make out the audio, it'll just be very clipped and noisy. That was the result when I did this with an RTLSDR, at least