Wizard-of-Oz (WoZ) is an essential step towards autonomous human-robot interaction. It enables simulating and modeling the interaction, moreover it allows the collection of data and experimenting without relying on specific system components. In child-robot interaction studies with the WoZ setup, the wizard speaks with the children through the robot. Thus, it is essential that the voice characteristics of the wizard fits the robots morphology. To ensure this fit, we explore the relation between the morphology of two robots (Probo and Nao) and human expected robots voice. To be able to investigate the fitting voice, we have generated variations of voice styles by using a voice modification architecture which provides global spectral shifts and vocal tract modifications. Our results confirm that indeed there is an obvious relation between the morphology and the voice pitch. The proposed voice modification architecture has been used in a data collection experiments with children interacting with Nao in aWoZ setup, where the result of the voice style study is applied on the voice of the wizard.