A vocal training apparatus accomplishes reproduction of a music signal recorded on an information medium, where the music signal includes a model vocal signal and an accompaniment signal. A model vocal signal is extracted from the music signal and a singer's vocal signal is received. The model vocal characteristics are measured from the model vocal signal and the singer's vocal characteristics are measured from the singer's vocal signal. The singer's vocal characteristics and the model vocal characteristics can then be compared to evaluate the singer's vocal characteristics relative to the model vocal characteristics.