Starting today, when you ask Alexa on your Amazon Echo speaker “what’s the latest?”, you may notice its voice sounds different – or more natural – than before. This is because Amazon has rolled out a fresh new speaking style to Alexa users in the US, modelled on the way newscasters deliver the news.
First announced in November 2018, Amazon’s developers utilized developments in Neural TTS (Text To Speech) technology to create the updated voice, which allows the voice assistant to tailor its speaking style to different commands and requests. In this case, Alexa’s newscaster-style voice has clearer intonation and emphasizes words in a way that sounds less robotic than its default voice.
To determine how important a natural-sounding speaking voice is to consumers, Amazon conducted a large-scale experiment, finding that people preferred Alexa’s newscaster voice when hearing the news.
However, subjects still preferred the voice of a human newcaster over Alexa’s new sound, suggesting consumers still feel more comfortable with voices that sound more like their own.
The researchers also experimented with creating a neutral voice for when Alexa reads out Wikipedia entries – but you can hear the different speaking styles via a Soundcloud embed over at TechCrunch.
Amazon also noted that researchers were able to train Alexa to use the newscaster voice after just a few hours of feeding it phonemes, – the most basic units of language that distinguish one word for another. (For example, the letters ‘p’ and ‘b’ that distinguish the words ‘pat’ and ‘bat’ are both phonemes.)
In its original statement in November, Amazon said that “this advance paves the way for Alexa and other services to adopt different speaking styles in different contexts, improving customer experiences.”
It’s easy to imagine the different situations this could come in handy, with one application being children’s bedtime stories. Although Alexa can already read your child a story – even personalizing it to include their name – how great would it be if it could read in dramatic voices or impersonate different characters?
Alexa can already lower its voice, when it’s late at night and you don’t want to disturb the people around you. If you enable Whisper mode, Alexa will respond to your hushed tones in a whisper – but like the newscaster voice, the feature is only available in the US right now, with no word on when it will arrive in other territories.