Fig. 3From: Developing a unit selection voice given audio without corresponding textSteps followed at synthesis timeBack to article page