This paper describes a Chinese text-to-visual speech synthesis system based on data-driven (sample based) approach, which is realized by short video segments concatenation.

To achieve the natural synthesized speech, the prosodic structure of input text should be precisely predicted by the text to speech Synthesis system.