Logo multitel

TTS-Voice synthesis

Text-to-Speech synthesis is used in many applications.
For instance, we use it where speech can replace or complement an interface, so that the machine can transmit information to a person. For example, in order to inform that a given specific application is starting.
Sometimes, speech is the only possible interface, for example, when the speaker cannot look at the screen to read the message (blind person).

The activities of Multitel in Text-to-Speech synthesis are focused on applied research, software development and consulting :

Applied Research

This activity consists in studying linguistic theories, computer-based and algorithmic principles, and mathematical bases and concepts required for developing TTS software. While remaining aware of new technologies, Multitel is currently interested in Non-Uniform Units (NUU) based synthesis, in spell-checking, and in automatic text-pattern recognition (sender, recipient, address, table columns,...).

Software Development

Constant research at Multitel in Text-to-Speech synthesis has led to the development of eLite. Based on the OOBP (Object Oriented Block Programming) paradigm, the eLite software is a complete Text-to-Speech synthesis system, wich provides a useful platform for the development and the assessment of new algorithms used in synthesis. eLite includes Eliot for the Natural Language Processing (NLP), whereas the Digital Signal Processing (DSP) is performed by either MBROLA, or the combination of a non-uniform units selector, LiONS, and a MBROLA-derived algorithm, TP-MBROLA.


Besides research and development activities, Multitel provides consulting in speech synthesis: help for software development (specification and prototyping), textual data collection for new languages designing, etc.