If you pick a bad enough codec, it should mask watermarks by distorting the sound, while still allowing speech to be audible. Maybe something like mu-law, whic

If you pick a bad enough codec, it should mask watermarks by distorting the sound, while still allowing speech to be audible. Maybe something like mu-law, which only has 3 kHz bandwidth centered on human speech (as opposed to mp3 with usually 44.1 kHz) would perform best.