Amazon’s Alexa voice assistant will soon be gaining a slightly upgraded voice, as well as a separate new voice that will be used for newscasts, according to The Verge. Alexa’s existing voice uses concatenative speech synthesis. The new voice, which is still recognizable as Alexa’s voice, uses neural text-to-speech (NTTS). The new NTTS voice uses machine learning to generate expressive voices more quickly.
The new newscaster voice was created by using machine learning to spot patterns in how actual newscasters read text. Amazon says that surveys show that customers prefer the new voices to Alexa’s existing voice. Have a listen for yourself.
Alexa’s Existing Regular Voice (Using Concatenative Speach):
Alexa’s New Regular Voice (Using NTTS Speach):
Alexa’s New Newscaster Voice (Using NTTS Speach):
Amazon says it only took a few hours to teach Alexa the newscaster voice. This could mean that Alexa will have many different voices in the future, each tailored for the type of text being read. You can expect the new NTTS Alexa voice, as well as the newscaster voice, to roll out to Alexa devices in a few weeks.
@AFTVnews
AFTVnews
I wonder how this will apply to how she reads Kindle books, because that could definitely use improvement.
Don’t they just play the pre-recorded audible version of the book? I think TTS of books was dropped when kindles lost their headphone jack.
No, they do the TTS. It’s horrible, which is why I don’t listen to Kindle books on my Echos.
Amazon’s in a weird spot here because if they improve Alexa’s ability to read Kindle books too much, there’s no incentive to pay for Audible. But it would sell more Kindle books! Will be interesting to see how this shakes out.
I remember that there was some speculation that Alexa was getting a new voice after the Super Bowl commercial. The voice at the very end of the commercial sounded slightly different and more natural, a lot like this new voice.
Will the new voice be rolled out to existing devices or only upcoming ones?