A method and a text subtitle decoder for decoding a text subtitle stream recorded on a recording medium are disclosed. A subtitle loading buffer initially loads a text subtitle stream including a dialog style segment defining a group of region styles and a dialog presentation segment having dialog presentation information and dialog text data for a dialog having at least one region. A text subtitle processor parses the dialog presentation segment into composition information, rendering information, and the dialog text data for each region. Next, a text renderer renders the text strings into a bitmap object for each region according to the rendering information, and the rendered bitmap object is composed in a graphics plane.