Effects of Number of Voices and Voice Type on Storytelling Experience and Robot Perception

Abstract

Social robots as storytellers have great potential in regards to replay every possible voice recording but also produce an endless variety of expressive synthetic voices themselves to illustrate different story characters. In an online study, we compared type of voice (human vs. synthetic) and number of voices to investigate the effects on storytelling experience and robot perception. Results show that recipients' transportation into the story was higher using a single voice compared to using different voices independent of type of voice. The same pattern was found for perceived robot intelligence. Anthropomorphism was higher for human than for synthetic voices. Further, animacy and likeability differed between type of voices in dependence of their number. No significant differences were found for warmth, competence, discomfort, or perceived safety. In general, when using human voice in robotic storytelling a single voice for the whole telling seems to be preferable. Further, when focusing on the storytelling experience in terms of transportation a single voice should be preferred, independent of voice type. Illustrating different characters by different voices is only suggested when utilizing synthetic voices. Overall, our findings support the claim that modern synthetic voice engines achieve results close to human speech and may have reached a point where they can deliver narration as credible as human voices.