In a study which is published as LNCS12284 82 participants investigated in a listening test via an online survey which synthesiser would be suitable as an expressive speech synthesiser for a Voice Output Communication Aid. eSpeak, MaryTTS, and CereVoice were tested annotated with SSML-tags. They had to listen to 3*3*3 phrases in happy, angry, or sad contours. They had to rate each on a 5-point Likert-scale. CereVoice scored better than MaryTTS than eSpeak.

?