Microsoft researchers have developed technology that allows software to use the sound of the user’s own voice to speech synthesize other languages.
During an open day of demonstrations run by Microsoft Research dubbed TechFest, Microsoft researcher Frank Soong demonstrated the results of software reading out other languages including Spanish, Italian and Chinese Mandarin in the voice of Microsoft research chief Rick Rashid.
While the results could hardly be called natural sounding, they do indeed sound rather like the source voice recorded in English. Microsoft said one application is for speech recognition and translation for travellers in foreign countries.
Microsoft also suggested the software could be used to help students learn a second language by providing phrases that would be easier for the student to imitate. The prototype software needs around an hour to train in order to be able function.
Another demonstration showed how the technology could also be used to supplement English speech synthesis so that foreign language remained consistent rather than switching to another voice for place names, for example.
A report on MIT’s Technology Review has a number of sound samples which demonstrate the technology while Microsoft’s TechFest keynote is also available in video here, assuming you have a Windows Media plug-in.