Google has come up with something called Google Translatotron. Science fiction movies are now coming true. Everything that you once deemed impossible is now being turned into reality. Google is also working actively towards taking this world to the future. The possibilities are endless.
Google has come up with a new technology which they like to call Translatotron. They say its an experimental new system. It will eradicate the need for any text and will directly translate speech to speech.
What is it called?
“Dubbed Translatotron, this system avoids dividing the task into separate stages,” was said in a blog post by the software engineers at Google AI named Ye Jia and Ron Weiss.
Google has identified a total of three stages of the translation systems that are in function today. One is automatic speech recognition which works on the principle of taking in speech and then transcribing it into text.
There is also machine translation which translates the transcribed text into the language in which the output is desired. The last one is called text to speech synthesis which uses the translated text to generate speech.
The integration of these systems created services that we fairly commonly use today, for e.g. Google Translate. Google is saying that this time around they will use a single model. They are trying to eliminate text completely from the process.
Google Translatotron: How does it work?
“Translatotron is the first end-to-end model that can directly translate speech from one language into speech in another language,” a Google AI blog post on Wednesday said.
Google says without a doubt that there will be an increase, a major increase in the speed of translation and there will be far less compounding errors.
They say it’s going to use spectrograms as input and generate spectrograms. It will rely on a neural vocoder and a speaker encoder. The system is going to keep the vocal characteristics of the speakers once the translation is done.
For more updates, stay tuned to The Geek Herald.