It's been one of science fiction's biggest dreams: a true universal translator. While we may be a long way from Star Trek's technology, a team at Microsoft Research is working on software that will enable a person's voice to be translated into other languages while still having the voice sound like the person who is speaking it.
MIT's Technology Review web site reports that this week the team, based at Microsoft Research Asia in Beijing, showed off their efforts at Microsoft's world headquarters in Redmond, Washington. The team used the voice of Craig Mundie, Microsoft's chief research and strategy officer, to greet an audience in Mundie's native English. The software was then used to have Mundie's words spoken in Mandarin Chinese, which still sounded as if Mundie had said the words.
Microsoft Research Asia's project needs about an hour at first to make a software model with the ability to read out any text in a person's own voice. Then the model is converted into another one that's able to read out text in one of 26 supported languages. This is done by comparing it with a regular text-to-speech model for the language that has been selected.