Humanoid robot is a robot that has intelligence like human. In this research, the team has build a robot called by FLoW. The robot is designed to have the ability as human beings, one of the ability is able to communicate. In the process of communication requires media, one of them is sound. This system is built to help the development research of ER2C (EEPIS Robotic Research Center) in building a Humanoid Robot 'FLoW'. Robot 'FLoW' to be able to communicate, then the robot should be able to say word or doing speech. Its called as speech generation. To generate sound, it will be make text to speech synthesis system. In the process of preprocessing is using FSA (Finite State Automata) algorithms. In Indonesian language uses 11 patterns. The testing process is done on the processing of 'words', 'sentences', and 'articles'. The percentage of success in 'words' and 'sentences' is more accurate and match with the separation of syllables in Indonesian language than the process of articles. From processing the article in newspaper, it has success rate of parsing 92.63%. The data were processed taken from five types of theme articles, namely the economy, education, sports, politics, and law. The performance is the result of parsing the articles is lower due to the addition of the name, title, and foreign words that have not undergone uptake in Indonesian language.