Amazon Introduces New SSML Tags to Make Alexa Sound More Human

Alexa might start sounding more human thanks to new SSML tags from Amazon

By Lulu Chang April 28, 2017

hiveypie — Image used with permission by copyright holder

You can almost convince yourself that Amazon’s Alexa is the best friend you never had. She’s always there to listen, she’s always there to help, and she’s always there with a response. But alas, sometimes the way in which those responses are delivered reminds you that she is nothing but an artificially intelligent voice. Amazon wants to help you forget that by making Alexa sound a bit more human.

On Thursday, Amazon announced a series of new Speech Synthesis Markup Language (SSML) tags that will help developers control Alexa’s intonations a bit better than before, making her sound more sophisticated — or at least, more like us. Those new tags have debuted in the U.S., the U.K., and Germany, and include whispers, expletive bleeps, and emphasis.

So what do the five new tags do? There’s whisper, which is quite self-explanatory — developers can use this tag when they want Alexa to engage in softer dialogue. As for expletive beeps, Alexa can now censor herself when she comes upon a naughty word. Developers can use the <sub> tag when they want Alexa to deviate a bit from her script, and the <emphasis> tag to change the rate and volume at which Alexa speaks. Finally, the new Prosody tag can be used to manipulate the volume, pitch, and rate of speech. Soon, Alexa will be pausing, whispering, and pretending to curse just like we do.

Of course, Alexa already has plenty of human-like qualities, or at least, qualities intended to make you believe she’s more human than she is. She can tell jokes, respond in kind to “I love you,” and get downright existential for the most philosophical of her users. But making her sound more like a person could go a long way in endearing her to you and your family.

Developers interested in learning how to make Alexa sound more like a real friend — which she isn’t — can join Amazon for a live webinar on SSML on May 18.

Editors' Recommendations