A “Babel Fish,” a small fictional creature that slips into your ear and translates any language in the entire universe, is one of the best inventions in all of sci-fi. Douglas Adams’ clever little device easily solved any language problems Arthur Dent might have encountered in The Hitchhiker’s Guide to the Galaxy. It’s also the first thing we think of anytime we hear about a company developing a device that might one day allow us to talk to anyone on the planet, even if we aren’t speaking the same language. Now Google is working on a new A.I. technology that won’t just translate for us, it will let us communicate in our very own voice.
Work on the Translatotron (which we learned about at Gizmodo) was announced on the Google AI Blog. Unlike other systems which, rely on three separate components (turning speech to text, translating the text another language, and generating that new text into different speech), Google is experimenting with what it calls a “new system that is based on a single attentive sequence-to-sequence model for direct speech-to-speech translation without relying on intermediate text representation.”
We’ll translate what that means. Unlike the current Google Translate, which goes from speech to text, and then from text to speech using a voice converter, this “end-to-end” method would turn your speech directly into another language without the middle steps, all while retaining your own voice.
Google provided real audio samples of how it works, though obviously there’s still a long way to go before this is a seamless conversion. But if perfected it would allow you to speak in English and be instantly heard in French, with your own voice (for good and bad), to someone else, just like a real Babel Fish.
Now we’ll be able to say to anyone, anywhere in the world, no matter what language either of us is speaking, “Doesn’t this Translatotron remind you of Hitchhiker’s Guide to the Galaxy?”
Images: ABC, Buena Vista