In this paper we describe a text analysis system that performs all the necessary steps to extract as much information as possible from unrestricted plain Spanish text, in order to give it as input to a units concatenation based speech synthesizer. We have developed a set of computer programs, in C language, to make a computer simulation of the analysis system and to evaluate its performance. This analysis system is a part of a Spanish TTS that is being developed by Telefonica I + D and AT&T Bell Labs.