Investigation into the Creation of Digital Voices,

1
Investigation into the Creation of Digital
Voices,

• with particular reference to the Festival System

Presented By M. Hood g01h0708 Supervised
By S. Bangay A. Lobb
2
Speech Synthesis has been around for years.

• First machines over 100
• year ago.
• Dedicated hardware has
• been available for 20 years.
• In Home PCs since the
• first sound cards.
• Latest software can almost fool a listener.

3
The Basics
Input

• Text Analysis
• Tokenize String
• Extract Syllables

• Linguistic Analysis
• Phonetic Analysis
• Utterance Generation

Spoken

• Waveform Generation
• Add emotional information
• Generate output from sound clips

4
Uses

• Internet avatars (news readers),
• Digital actors,
• Game characters,
• Next generation chat applications
• Systems typically only have two voice,
• One male, one female,
• Other voices only variations on these.
• Limits application when many characters on screen
  at once.

5
The Festival System

• Open source Speech Synthesis package, developed
  at the Centre for Speech Technology Research.
• Based on the Edinburgh Speech Tools.
• Supports multiple voices, in multiple languages.
• Allows new voices and languages to be added by
  use of a scripting language similar to Lisp.
• FestVox.
• It is proposed this project investigates speech
  synthesis software, and examines the issues
  involved in the creation of additional voices for
  Festival.

6
Project Plan and Time Frame

• First Term
• Research other papers in digital voice creation.
• Experiment with various speech synthesis tools.
• Learn and become proficient with Festival.

7
Project Plan and Time Frame

• Second Term
• Create a number of voices from scratch.
• See how time spent on the voice affects the
  quality of the voice.
• Create some voices by editing current voices.
• Create a new language.
• Import voices from one language into another.

8
Project Plan and Time Frame

• Third Term
• Work on the creation of an automation tool.
• Look into adding expressions by adding sound
  filters to the voice.

9
Project Plan and Time Frame

• Fourth Term
• Tie up loose ends.
• Write up.