An audio output apparatus including a digital operation unit, a digital-to-analog converter (DAC), a left channel unit, a right channel unit and a common unit is provided. During an initial setting, the digital operation unit detects a cross talk voltage on a left channel earphone unit in the left channel unit or a right channel earphone unit in the right channel unit, calculates a first ratio and a second ratio, and then performs an arithmetic operation on a left and a right channel signals according to the first and the second ratios, so as to obtain a compensated left channel signal and a compensated right channel signal respectively for eliminating a cross talk phenomenon.