Place, authenticity time: a framework for synthetic voice acting

Abstract

Current solutions to computer speech are evaluated by the degree to which the synthetic speech signal replicates a human speaker. While there has been significant progress in this area, users are still unenthusiastic about most synthetic speech. The Place, Authenticity, Time (PAT) Framework provides an alternative perspective that addresses the problem of ‘likeabilty’ for synthetic speech. The multidisciplinary framework draws upon concepts of ‘liveness’ from performance theory, theories of spontaneous speech from linguistics and techniques derived from performance practice including renaissance verse speaking. This article presents an outline of the theoretical position and a summary of findings from user evaluations: that users will be ambivalent about the truthfulness of a human or synthetic voice; that what a voice says is important; how it says it is also important; humour may be helpful in achieving user acceptance and the setting of the voice and user is also significant.