Real time voice translation (Think Babelfish from Hitchhikers Guide or the Universal Translator in Star Trek) is something that up until now has been science fiction.

Once ready, the service will work just like a real professional human interpreter.  It will analyze packets of speech instead before it attempts to translate since if you’ve ever studied a foreign language you know that word for word translation will frequently result in incomprhensible sentences.

However, Google is working hard to pioneer real time speech to speech translation and hopes to have a working service up in the next few years.

Google can already do a pretty good job with translating web pages into different languages.  However, machine translation of text is much, much easier than voice translation.

The two main challenges right now are doing high accuracy machine translation and high accuracy voice recognition.

The machine translation part will probably be ready for prime time long before the voice recognition portion.

If you’ve ever used Google’s Visual Voicemail service (free with your Google Voice account) you know that voice recongition definitely has a ways to go.  It can transcribe clealry spoken Enlish from a Native speaker fairly well, but anyone with an accent tends to be mostly gibberish.

Since each person has a different voice, accent, and pitch, true voice recognition is extremely difficult.

However, if you’ve ever used an Android phone, you know Google is working hard to make voice recognition a reality.

Their smartphone operating system can do a lot with voice searches and they even have a program you can download from the Android Marketplace called “Talk To Me.”  It’s definitely limited in what it can do, but it’s definitely a cool program to play with if you need translation (both text and voice) in a pinch.

While the technology may someday support a significant portion of the world’s 6,000 plus languages, expect support for popular languages like Spanish, French, German, Japanese, and Mandarin to be among the first supported.