Speech synthesis devices

PURPOSE:To provide a synthesized voice of high quality by providing an option function when plural different synthesis systems such as rule synthesis, analytic synthesis, and sound recording and reproduction are combined. CONSTITUTION:A document which is inputted from a document generation part 1 is added to a document to be sent by a rule synthesis part 3 by a part 2 which determines a synthesizing means according to the document contents and 2 or 3 characters of a following character string are sent together. The document sent to the rule synthesis part 3 is analyzed by a language process part 3a and a rhythm control part 3b into reading, accent, and part-of-speech information, a parameter generation part 3c generates voice parameters, and a synthesis processing part 3d generates a synthesized voice. Further, the document sent to a recording and reproduction part 4b has matching voice waveforms extracted by a waveform data storage part 4a and a read control part 4b respectively. An overlap process part 5 adds a waveform to the part where waveform data for rule synthesis and waveform data for sound recording and reproduction overlap with each other and the result is outputted as a voice from a synthesized voice output terminal 7 through a D/A part 6.