The present disclosure provides a method for adding realism to synthetic speech. The method includes receiving text (218) that is to be converted into synthetic speech from a mobile device (108). The text (218) may include embedded emoticons indicating a first prosody information and a predefined sound stored in a stored data repository (208). The method also includes identifying a user associated with the text (218) based on a comparison between metadata associated with the text (218) and user profiles stored in the stored data repository (208); retrieving a speech font from a speech data corpus associated with the user stored in the stored data repository (208). The speech font includes a second prosody information and a predefined accent of the user. The method further includes converting the text (218) into synthetic speech based on the retrieved speech font, which is being modulated based on the emoticon.