This paper describes how the SPRUCE (Speech Response from Unconstrained English) system is capable of acting as a "front end" to any text-to-speech system whether it be based on diphones, phonemes, demi-syllables, syllables or even words. This is possible because the architecture of SPRUCE includes a large dictionary which contains phonetic information of individual words along with syllable, stress and other prosodic information. This representation is such that the basic linguistic information for driving any text-to-speech synthesiser is already contained within the SPRUCE synthesis system. The advantage of using SPRUCE as a front end for a synthesiser is that it enables the synthesis system to make use of SPRUCE'S large word dictionary and its natural sounding intonation algorithm.

Keywords: SPRUCE, text-to-speech synthesis, syllables