Google introduces Translatotron, a Speech-to-Speech translation system

Google introduces Translatotron, a Speech-to-Speech translation system
Google introduces Translatotron, a Speech-to-Speech translation system

Google has declared its initial guide speech-to-speech translation system known as “Translatotron” that may convert verbal communication from 1 language to another while keeping up the speaker’s voice and speed.

“Translatotron” relies on a sequence-to-sequence system that takes resource spectrograms — a visual representation of frequencies as input and generates spectrograms of the translated content from the target language, Ye Jia and Ron Weiss, applications engineers in Google Artificial Intelligence (AI) wrote in a blog article on Wednesday.

The model uses two other individually trained parts — a neural vocoder which transmits output spectrograms into time-domain waveforms plus a speaker encoder which may be employed to keep the character of the source speaker’s voice from the synthesized interpreted language.

For the time being, the outcomes of Google’s demonstration of this translation system lag behind a traditional cascade program, but the search engine giant said it’s “demonstrated the feasibility of this end-to-end speech-to-speech translation”.

“By incorporating a speaker encoder network, ‘Translatotron’ is also able to retain the original speaker’s vocal characteristics in the translated speech, which makes the translated speech sound more natural and less jarring,” Jia and Weiss explained.

The business stated that”Translatotron” is your very first end-to-end model that may directly interpret speech from 1 language into another.

“We hope that this work can serve as a starting point for future research on end-to-end speech-to-speech translation systems,” that the blog-post noted.

Recently, Google has been working aggressively on its own translation systems.

Adding more languages into its real-time translation attribute, sooner in 2018, the business introduced beams in Google Translate with many different languages from region-based pronunciations.

Also Read: Mobile ticketing firm TodayTix collects $73 million to improve its Services in App

Back in February, Google expanded the “Interpreter” manner on its Assistant to Home speakers along with other third-party apparatus to assist users to have real-time discussions with compatible devices and interpret them into 26 recorded languages.

Google’s yearly developer conference, I/O 2019, kicked off with large upgrades to its crucial services, including Google Search, Assistant on Android. Google introduced Pixel 3a collection. The newest Pixel is introduced beneath”cheap” flagship mobiles.


About the author


Alex Penfold


Alex is Senior Content Writer at TechRazr. He is being writing news for various blogs around the Internet. He likes to test latest app in the market and share with users. Also, he loves travelling.

To get in touch with Alex for news reports he published you can email him on [email protected] or reach him out in social media linked below.

Add Comment

Click here to post a comment