In this project we are using Tacotron to build text to speech for Tamil language. First we are changing the configuration to train the model for Tamil language then by giving the sample text in Tamil language it will give the speech format output. Due to less configuration system we not able to train the model with huge amount of data. For now we trained the model with 12k records so it is not properly generating the speech of the given tamil text. Our next goal is to train the model with high-end congiguration system and then fine-tune the audio to get the proper speech of the given text.