Voice servers are now available for Public Switched Telephone Network subscribers. These devices make use either of stored announcements or of synthetic speech. Synthetic speech may be produced from stored segments such as words, syllabes or diphones; it may also be produced by synthesis by rule, e.g. formant synthesis. In all cases of signal processing, such as digital compression of the signal, together with sound processing such as concatenation of segments and variation of pitch, intensity and segment duration, a noticeable impairment of speech quality may occur.