Mobile communication is now shaping our daily lives. However, as the end-user devices become increasingly smaller, the limited screen size and keyboard space start to bring about lots of inconvenience. Therefore, these conventional input/output methods are challenged by more natural interfaces, such as voice interface. Speech synthesis, which is one important aspect of voice interface, has so far been successfully applied to mobile phones to read email, instant messages for us. Nevertheless, these text-to-speech (TTS) systems, which are primarily based on concatenation of acoustic units, usually suffer from audible discontinuities between two successive acoustic units. This paper presents a novel method that can effectively remove the audible discontinuities in the Chinese speech synthesis and highly maintain the speech naturalness.