Can't use an NPN for this! It would only conduct the positive side of the waveform. That signal probably has an output cap, so it's just distortion.
The proper solution for this job is an analog switch such as the 4051, 4052, or 4053. It is a bidirectional switch. The switching input is very high impedance so you're free to design the driver without a current problem.
The problem with this is usually getting the sound to turn off fast enough; ideally you want the sound to have cut off a moment before the priority sound source starts talking but that's usually not possible.