The research prototypes and early models that come out of Microsoft’s research department never cease to amaze. Microsoft Research in Asia has recently unveiled a new speech recognition technology at TechFest 2012 that can, after fine-tuning it to your voice, can translate it recognizably into other languages.
Understanding that accent and how you say things is important, the program keeps accurate speech in a consistent voice. The program builds a model of any voice, so upon programming or specializing it to your voice, it can effectively “speak” for you in other languages. They also showed with all this that they were able to make a 3D model of a person which can conform to apparently speak the translation, likely after yet more fine-tuning.
Keep in mind that this is, of course, a prototype, but tit can already do some pretty impressive stuff. Don’t forget to check out Microsoft Research’s video with the Microsoft reps showing off the new toy about 15:00 minutes in, though the full keynote starts at just after 12:00 minutes.
Source: Extreme Tech