Building Voices in

the Festival Speech Synthesis System

Processes and issues in building speech synthesis voices

Edition 1.0, for Festival Version 1.3.1

24th January 1999

Full distribution with document sources and related scripts festvox.tar.gz
Postscript version of this document alone festvox.ps.gz

NOTE: this document is incomplete

• 1 Abstract
• 2 Copying
  • 2.1 Acknowledgements
• 3 Introduction
  • 3.1 Text-to-speech process
    • 3.1.1 Text analysis
    • 3.1.2 Linguistic analysis
    • 3.1.3 Waveform generation
  • 3.2 Requirements
    • 3.2.1 Hardware/software requirements
    • 3.2.2 Voice in a new language
    • 3.2.3 Voice in an existing language
• 4 Utterances
  • 4.1 Utterance structure
  • 4.2 Modules
  • 4.3 Utterance access
    • 4.3.1 Features as pathnames
    • 4.3.2 Access idioms
  • 4.4 Utterance building
  • 4.5 Extracting features from utterances
• 5 Diphone databases
  • 5.1 Diphone introduction
  • 5.2 Defining a diphone list
    • 5.2.1 Synthesizing prompts
  • 5.3 Recording the diphones
  • 5.4 Labelling the diphones
  • 5.5 Extracting the pitchmarks
  • 5.6 Building LPC parameters
  • 5.7 Defining a diphone voice
  • 5.8 Checking and correcting diphones
  • 5.9 Diphone check list
• 6 Unit selection databases
  • 6.1 Cluster unit selection
    • 6.1.1 Collecting databases for unit selection
    • 6.1.2 Preliminaries
    • 6.1.3 Building utterance structures for unit selection
    • 6.1.4 Making cepstrum parameter files
    • 6.1.5 Building the clusters
    • 6.1.6 Defining a voice
  • 6.2 Diphones from general databases
• 7 Building prosodic models
  • 7.1 Phrasing
  • 7.2 Accent/Boundary Assignment
  • 7.3 F0 Generation
    • 7.3.1 F0 by rule
    • 7.3.2 F0 by linear regression
    • 7.3.3 Tilt modelling
  • 7.4 Duration
  • 7.5 Prosody Research
• 8 Lexicons
  • 8.1 Word pronunciations
  • 8.2 Lexicons and addenda
  • 8.3 Out of vocabulary words
  • 8.4 Letter to sound rules
  • 8.5 Post-lexical rules
• 9 Text analysis
  • 9.1 Token to word rules
  • 9.2 Number pronunciation
    • 9.2.1 Multi-token numbers
    • 9.2.2 Declensions
  • 9.3 Homograph disambiguation
  • 9.4 TTS modes
• 10 Full example
• 11 Evaluation
  • 11.1 Does it work at all?
  • 11.2 Sematically unpredictable sentences
• 12 Resources
  • 12.1 Festival resources
  • 12.2 General speech resources
• 13 References
• Index

This document was generated on 11 Febuary 1999 using the texi2html translator version 1.52.