CMOS_stylettsz_flashspeech_librispeech

1. To participate in this survey you must be a native English speaker located in the U.S. The survey will take around 10 minutes to complete.

You will be listening to 20 sets of audio samples. In each set, there is one reference voice and two audio recordings (A and B) of the same text. Some audio is synthesized while others are read by humans. Please listen to the audio clips carefully, compare each audio with the reference, and then

1. Rate how natural is B compared to A on a scale of -6 to 6, with 6 indicating that B is much better than A. Here, naturalness includes whether you feel the speech is spoken by a native English speaker from a human source, as opposed to being synthesized by a computer. Please note that the reference speech is spoken by a human, so you may use that to judge naturalness.

2. Rate how similar is the speaker in B to the reference voice, compared to A. Here, "similar" means that you feel that the recording and the reference voice are produced by the same speaker.

There are some audios for checking your attention through the survey. If a certain number of the scores are rated unreasonably, you will not be paid. So, listen to the audio carefully and do not fill out the survey randomly. *This question is required.

I have read the above purpose of the study, and understand my role in participating in the research. I volunteer to take part in this research. I understand that I may refuse to participate or withdraw from participation at any time. The investigator may withdraw me at his/her professional discretion. I certify that I am 18 years of age or older and a native english speaker and freely give my consent to participate in this study.

CMOS_stylettsz_flashspeech_librispeech_0519