TTS should not be confused with voice response systems. Voice response systems synthesize speech by concatenating sentences from a database of prerecorded words and are used for different purposes than TTS systems, which form sentences and/or phrases based on a language’s graphemes and phonemes. Voice response systems are limited to synthesizing sentences that contain only words that have been predetermined by the system. TTS systems, in contrast, are theoretically capable of "reading" any string of text characters to form original sentences.

A Short Introduction to Text-to-Speech Synthesis
An excellent overview of TTS systems.