October 23, 2019

Human, but not human, the voice of EMAR

Finding the perfect voice for EMAR is difficult. 

Last spring, we tested several robotic and human voices with teens and discovered two human-like voices originally stood out as ideal. Both were human voices recorded by Project EMAR students, both of which had a Chinese accent. We added a slight echo to make them less human. Originally a group of teens preferred these voices over robotic ones and said the accents were “soothing.” Until…

 

In a design session this past summer, we tested EMAR’s female voice with a group of teens on campus. A few teens in the group did NOT like the female voice in particular. One Chinese teen said it reminded him of his mom and that was “creepy.” 

 

So we headed back to the drawing board. HCDE Master’s student, Stefanie Gueorguieva continued the task as voice designer and determined the ideal voice be almost human, but not fully human, and maybe not a specific accent.  Stefanie compiled over ten voices that included a mix of online generated and original voices adapted using Audacity software. Each voice was unique and had different degrees of echo to determine if a more human or robotic voice would convey empathy towards teenagers. She found teens felt that voices with a smaller degree echo that seemed more “human” like conveyed more empathy than voices with a greater amount of echo or voices that were entirely computer-generated. Not surprisingly, teens also mentioned that they wanted to be able to customize the voice of EMAR to their liking. This aligns with the many other customizable aspects of EMAR.