I simulated the input transistor circuit. Its max gain is about 32dB. In the simulations I changed the scale so that max output is about 0dB.
1) With a 100nF input coupling capacitor almost all of the audible range is passed with -3dB at about 120Hz.
2) With a 10nF input coupling capacitor most voice frequencies are passed with -3dB at about 1200Hz.
3) With a 1nF input coupling capacitor the upper audible range is passed with -3dB at about 15kHz.
I think a "puff" is high audio frequencies. Then i would use a 1nF or 2nf input coupling capacitor so that normal speech, background sounds and music does not trigger the circuit.