I didn't find it too convincing either, although part of that could be priming with the initial unmodulated audio.

I'd be interested in a standalone study of how effective the modulation is at masking.