Abstract: In a telecommunications relay system employing automated text-to-speech for conversion of a text message entered by a sound impaired person, the delay between the completion of the automated text-to-speech conversion and the assignment of a live shared attendant for transcribing and transmitting that which the unimpaired person says is eliminated by completing the text-to-speech conversion for each text message and its associated end-of-message code word without the introduction of any delay and, immediately thereafter, accepting and storing the unimpaired person's speech into a voice buffer at least until a live attendant is assigned to the relay session. When an attendant is eventually assigned to the call, the attendant retrieves the speech stored in the buffer.