Text-to-speech

Speech synthesis is one of innoetics key technologies.

Our text-to-speech technology is based on state-of-the-art algorithms and techniques that allow the creation of synthetic voices from natural recordings, by maintaining at the same time the voice characteristics of the original speaker. It incorporates several sophisticated modules, such as text processing module and prosody module, which by combination they produce a near-natural synthetic speech from any given text.

The overal speech quality is defined by a set of criteria which they also depend on the context of application and the target area of use. The level of intelligibility and naturalness increases as the domain area is limited. Extreme cases could be for example weather reports where the dictionary is very limited.

The quality of a speech synthesizer is based on a four-fold criterion:

• naturalness
• intelligibility
• text processing
• appropriatness for specific application context

While the first two ccriteria refer mainly to the actual speech signal and its quality, the third one addresses the efficiency of the system against written text, like numerals, abbreviations etc. Depending on the needs of an application area, the aforementioned criteria can be prioritized differently, assigning a different grade to similar TTS systems.

Innoetics offers the beyond state-of-the-art text-to-speech technology for the Greek language. Design and development of our systems have been perfomed in a user-centric manner for best results against all kinds of challenges.