Of primary interest in any method for producing synthetic speech is to minimize the number of bits per second required to generate acceptable quality speech. An efficient method for transmitting the linear prediction parameters has been obtained by using the techniques of differential PCM. Using this technique, speech transmission employing fewer than 1000 bits/sec is achieved. Further reductions in the linear prediction storage requirements can be realized at a cost of higher system complexity by the transmission of the most significant eigenvectors of the parameters and a form of optimal pre-emphasis. These techniques in combination with differential PCM can lower the storage to 600 bits per second. Tape recordings of the generated synthetic speech will be played.