Speech
Research Laboratory
AI. duPont Hospital for
Children
and the
University of Delaware
Info for Users (text)
ModelTalker Speech Synthesizer
The ASEL ModelTalker TTS system converts plain English text to speech. It uses a text to phoneme system which includes capabilities for parsing ToBI-like descriptions of the intonation. Synthesis is accomplished through a combination of database-driven speech and a variant on diphone-based phoneme to sound engines known as Biphone-Constrained Concatenation (BCC). Speech stored in the database encompasses common words and phrases in different contexts as well as a complete set of biphones. The BCC sound engine results in smoother, more natural speech, without sacrificing the ability to quickly "capture" new voices in the biphone inventories for the system. Resulting synthesized speech sounds like, or model, the speech of particular talkers with, at its best, the quality of recorded speech, or, at its worst, the quality of synthetic speech in the same "voice".
Download InvTool Speech Recording & ModelTalker Speech Synthesis Software
Examples of synthetic speech:
Synthesized speech found entirely in speech database:
2) What do you think about that?
Synthesized speech with both long and short stretches of speech from database:
3) I left my coffee on the table.
4) The storm is getting worse outside.
Synthesized speech composed almost entirely of diphones in database:
5) She touted the virtues of Guar Gum.
6) The obstreperous youth broke the carousel.
New voices and our synthesizer will be online shortly!