March 2, 2010
ICSI Speech Group alum Matthew Aylett has developed text-to-speech software that allows film critic Roger Ebert, who lost his voice box four years ago, to speak in an approximation of his own voice. "It still needs improvement," said Ebert in a report aired on CBS, "but at least it sounds like me." Aylett's Scotland-based company, CereProc, broke down voice recordings from Ebert's film commentary into individual sounds, which are reassembled when Ebert types a sentence into the software. Ebert can also adjust the emphasis and intonation of the synthesized sentences. CereProc offers a number of other voices produced from text, including voices with Irish, Scottish, Southern English, and American accents. Ebert's new voice debuted on the Oprah Show in March.