In loudness normalization, voice dialogue usually uses 0 dB as the peak reference, and background music is pressed down to around -12 dB in order to widen the dynamic gap between dialogue and score and prevent the BGM from overpowering the lines.