A real-time system which translates any Italian text into speech having good naturalness and high intelligibility is presented. The system, which is planned for use in a reading machine for the blind, is built up using a commercial microprocessor system and a voice synthesizer chip. It is the present implementation of the system previously described in (ref 1,2). The text-to-speech conversion is carried out in a two-step procedure. First a linguistic analysis is applied on the input text in order to resolve the lack of one-to-one correspondence between graphemes and "phonemes", to assign internal word stress hierarchy and to perform some form of rudimentary syntactic analysis. As a result the text is converted to an abstract linguistic representation consisting of phonemes, stress marks and syntactic structure indicators. Then the sequence of phonemes is converted to sound by computing the proper prosodic functions (fundamental frequency, duration and intensity) to be applied to the set of linear predictive coded diphones used for synthesis.