Models of speech production and how these are used in text-to-speech conversion are reviewed. In the first part of the paper the foundation is laid for an explanation of present day speech synthesisers, and their limitations, through a phonetic description of speech production. The paper then presents a theorectical model of speech production which is the basis of most synthesisers. Next, a number of speech synthesisers are surveyed and their relative merits and shortcomings are considered. The paper ends with a brief look at techniques that are being used in place of the more traditional models of speech production in text-to-speech (TTS) systems and considers possible areas of progress in the future.