Last reviewed: 5/1/2010
Article ID: H051001
HOW: Synthesizing phonemes
The information in this article applies to:
- LexiconKit 3
- SpeechKit 6
- Chant Developer Workbench 2008
Word pronunciations are comprised of phonemes. These pronunciations are stored in lexicons to improve recognition accuracy and enhance synthesizer quality of synthesized speech. If you want to synthesize phonemes to hear how they are pronounced by a voice, then you must indicate the text includes phonemes on the playback.
LexiconKit provides facilities to generate, edit, and persist word pronunciations in libraries referred to as lexicons. Lexicons can be used by recognizers at runtime to help improve recognition accuracy. Lexicons can be used by synthesizers to help improve the quality of synthesized speech.
Chant Developer Workbench provides an editing surface for managing lexicon pronunciations. See Knowledge Base article HOW: Generating, editing, and compiling Nuance Vocalizer and Vocon user dictionaries.
In SpeechKit, you can synthesize text with phonemes for SAPI4, SAPI5, Cepstral, Nuance RealSpeak, and Nuance Vocalizer voices. Phonemes are unique to each speech API. See LexiconKit documentation for details of how to generate phonemes for recognizers and synthesizers. To synthesize phonemes in SpeechKit, use the CPSSpeakPhonemes playback style in your StartPlayback method as follows:
ChantTTS.StartPlayback("t$me&Ir6o&U", 0, CPOText, CAFDefault, "", CPRMultiMedia, CAFDefault, CPSAsynchronous + CPSSpeakPhonemes)