Skip to Content

Department of Linguistics

TTS Input: Plain text and other input formats

Robert Mannell

Plain Text

Traditionally, TTS systems took as their input simple ASCII text strings. These strings could be in the form of single sentences or longer samples of text, such as paragraphs or even longer text. The text could be submitted by typing at a terminal or by submitting the name of a text file. The older versions of the SHLRC TTS system took as input, a single sentence, submitted either by typing or by reading a text file. MU-Talk can handle any length of plain text (limited by available memory) and uses punctuation to determine sentence and paragraph boundaries.

  • For more information on the type of input handled by MU-Talk, click here.

Word Processor Formats

As TTS systems have been adapted to the task of dealing with real text, especially as it occurs on computers, they have needed to deal with much more complex input. For example, word processor text often includes large numbers of codes which must be ignored if a TTS system is to read only the text content and not other types of information (formatting information, for example).

Web Pages

TTS systems designed to work with Aural Browsers must read the meaningful text and mostly ignore the html codes, except when those codes are needed to help with emphasis or to produce changes in voice quality or speaker that indicate to vision-impaired users various aspects of web page layout.

  • For more information on TTS and web pages click here.

Text Generation System Output

TTS-generated speech can be used as the output for text generation systems. Such systems may, for example, provide speech output for database query systems.

  • For more information on TTS interfaces to text generation systems click here.

TTS Mark-up Languages

A number of mark-up languages have been developed that are used to control the way a TTS system handles input text. These mark-up languages vary in their goals. Some are high level languages used to control a limited number of speech and speaker variables, whilst some are designed to facilitate the detailed control of how the TTS system processes text. This category of input overlaps with item c), above, as a couple of proposed standards deal with mark-up for web pages.

  • For more information on TTS mark-up languages click here.