TTS (Text-To-Speech) systems generate speech equivalent for the text given as input. Though generation of speech is with moderate complexity the aspect of introducing naturality with the expression of the speaker is a big challenge faced in TTS. When in Tamil, because of its phonetics, the algorithms used for other languages like English will not work because of adaptive pronunciations which are completely dependent on the language constructs. Our proposed work will be in the TTS area of signal processing. Current researches have used HMM (A Hidden Markov model) as the machine learning algorithm for classification and is one of the first of the Indic word recognition system. In this paper, an HMM speech synthesis is done using the HMM for the Tamil language.