JUst CONcatenation - A Corpus-based Approach and its Limits

This paper describes a radical corpus-based approach to speech synthesis. No signal manipulation is performed and the synthesis becomes a mere concatenation. The feasibility of this approach is evaluated regarding corpus selection constraints and realization of different prominence patterns. A “traditional” concatenative system serves as a baseline. The results indicate that the size of the corpus must be rather large in order to obtain satisfying and reliable results for unlimited text-to-speech conversion.