Institute for Natural Language Processing

Speech Synthesis at the IMS

Speech Synthesis at the IMS

TTS demo

We provide two demo pages. On the first one you can synthesize any given text online. The second is a speaking metro information system for the region of Stuttgart.

Project description

The speech synthesis activities at the IMS (Experimental Phonetics) concentrates on various linguistic and application oriented aspects of speech synthesis. Our goal is to achieve naturally sounding and linguistically motivated speech synthesis.
The speech synthesis system developed at the IMS is the IMSGermanFestival system (download IMS German Festival). It is based on the original Festival speech synthesis framework developed at CSTR, University of Edinburgh. The current voice of our system uses diphones taken from the MBROLA project.
We are currently working on various aspects that will improve our system. One main aspect is prosody generation. We are exploiting methods that take into account linguistic information (like part-of-speech or syntactic parsing) to predict accents and phrasing. A sentence's pitch contour is generated using statistic or rule-based methods. Our speech synthesis system has a refined text-preprocessing module that deals with all kinds of abbreviations and various different number formats (cardinal and ordinal numbers, dates, currency, ...).
Our group is participating in the SmartKom project, a large German-wide project on multimodal human-machine interaction founded by the German gouvernment. In this project we are responsible for the speech output modality. Additional funding comes from the Deutsche Forschungsgemeinschaft (DFG) and the Sony Research Center Germany.

Members of the IMS speech synthesis group

• Martin Barbisch
• Grzegorz Dogil
• Bernd Möbius
• Bettina Säuberlich
• Antje Schweitzer

Former members of the IMS speech synthesis groups

• Norbert Braunschweiler (now with Toshiba, UK)
• Martin Haase
• Tanja Klankert
• Gregor Möhler (now with IBM, Böblingen)
• Edmilson Morais da Silva (now with Univ. Campinas, Brazil)

Download

You can download a version of our German Festival synthesis system.

Interesting local links

• A collection of text-to-speech systems with sound examples for numerous languages.
• Teaching speech synthesis. A site of the ESCA Special Interest Group in speech synthesis (SynSIG) dedicated to everything useful for teaching this subject (organisations - courses - software - tutorials - links - images).
• Linguistics & Phonetics Worldwide.

top