This is the documentation for the German version of the Festival speech synthesis system (http://www.ims.uni-stuttgart.de/phonetik/synthesis/). It was developed at IMS (Institute of Natural Language Processing) at the University of Stuttgart and is based on the original (mainly English speaking) Festival system (release 1.4.1) developed at CSTR ,University of Edinburgh (http://www.cstr.ed.ac.uk/projects/festival). This documentation describes modules that enables Festival to speak German. It can be seen as an extension to the original Festival documentation available at http://www.cstr.ed.ac.uk/projects/festival/manual/.
All modules that are provided on an open-source base are described here. Additionally we describe methods that are currently under investigation and, therefore, not included in the open source version of the IMS German festival.
The original Festival distribution was extended by a number of additional modules that are required for German. For instance, we have added a complete German text preprocessing that can deal with numbers and abbreviations in German. The same holds for various approaches of intonation generation. In other cases we have simply modified modules available for English. The duration module is one of these: It uses the same engine as for English, but with a different phone set, default phone durations and a CART tree trained on a German speech corpus. Other modules, like the MBROLA backend, were not modified.
This documentation, like most documentations, is growing. During the process of perfecting the synthesis system other modules will be added and existing modules will be modified. Therefore, expect this document to be updated from time to time.