HIGH-QUALITY TVML CONTENT BY USING POST-RECORDING PROCESSING AND EMOTION IN SPEECH

摘要

In past research, we developed a language called TV program Making Language (TVML) [1] for describing TV-like video content and software called TVML Player [1] for interpreting TVML and generating a TV program. We also devised a method that separates the program script from program direction so that even users with no video production skills could produce a TVML-based TV program by merely writing a simple script. The TV Creator [2] tool that we developed incorporates that method, but detailed program directions cannot be added with TV Creator alone. With the aim of enhancing the quality of TVML content (the "TVML program"), we have developed a post-recording process that can easily replace the synthesized speech spoken by computer-generated characters with natural human voices. And, in conjunction with this process, we have devised a method for automatically adding program directions such as for facial expressions and behavior of computer-generated characters, and constructed a trial application to assess this method. This paper describes the TVML-program production method and the challenges that it presents, and an application for performing post-recording processing that incorporates a method for automatically adding program directions to a TVML program.