To provide a device and a method for converting text voice in which high quality voice is outputted based on arbitrary text data.

In the device and the method, text data that are objects of voice outputs are inputted, the text data are collated with voice data beforehand stored, synthesized voice is generated by connecting the voice data outputted as the results of the collation and the generated synthesized voice are outputted. The collation is successively made from the voice data that are stored in a large voice unit. When the voice data are completely made correspond with the text data in a sentence unit, the voice data are outputted as a collation result. When the voice data are partially made correspond with the text data, the voice data, which are stored with a largest voice unit among the partially corresponded voice data, are successively outputted as a collation result.