To make the goal of building voices in new languages easier and more accessible to non-experts, the combined tasks of phoneme set definition, text selection, prompt recording, lexicon building, and voice creation in Festival are now integrated behind a web-based development environment. This environment has been exercised in a semester-long laboratory course taught at Carnegie Mellon University. Here we report on the students' efforts in building voices for the languages of Bulgarian, English, German, Hindi, Konkani, Mandarin, and Vietnamese. In some cases intelligible synthesizers were built from as little as ten minutes of recorded speech.
Cite as: Kominek, J., Schultz, T., Black, A.W. (2007) Voice building from insufficient data - classroom experiences with web-based language development tools. Proc. 6th ISCA Workshop on Speech Synthesis (SSW 6), 322-327
@inproceedings{kominek07_ssw,
author={John Kominek and Tanja Schultz and Alan W. Black},
title={{Voice building from insufficient data - classroom experiences with web-based language development tools}},
year=2007,
booktitle={Proc. 6th ISCA Workshop on Speech Synthesis (SSW 6)},
pages={322--327}
}