Today we're announcing a whole new set of technologies called neural text to speech.
This uses deep learning models to be able to model hurt boys in a much more natural and fluent way.
Best way is to kind of hear this because the pace of innovation is pretty crazy.
We took two clips.
The first clip I'll play is Alexa, when we launched her Five years ago, let's go ahead and play that.
Star Trek as an American media franchise based on the science fiction TV series created by Gene Roddenberry.
This blew people's minds when we launched it.
But let me tell you what we've launched into production as of today.
Why don't we play the clip using neural networks.
Star trek is an American media franchise based on the science fiction television series created by Jim Ladin Barry.
The pace of innovation is incredible, and what we can do with machine learning really nevcer cease to amaze me.
The same neural network technology that gives us a lot more flexibility now in and around what we can do wth Alexa's voice.
And so today we're also going to be rolling out multi lingual mode.
And this gives customers that are in bilingual households access to Alexa.
There are billions of how.
Households around the world that have dual speakers, sometimes three languages in a single household.
And we want Alexa to work great in that.
So now starting in the United States, customers will have access to Spanish and English, in Canada, French and English, and in India, Hindi and English.
And she will automatically switch between these language and her new models voices sound great in both languages.
Beyond language, that same neural text to speech allows us to let Alexa be much more emotive and expressive.
And we've modeled this in a bunch of different ways.
She can be excited when your team wins say the Cubs win not happening much right now, but if they did win, she can be expressive in a newscaster kind of way.
So we have a newscaster voice when she reads the news, you'll start hearing her cadence sound much more like a newscaster.
[UNKNOWN] Can mimic expressions that we've grown accustomed to.
I grew up listening to Sunday on the Radio, and it was always a case of counting down the top 40, and she can actually do that too, and mimic a little bit like a DJ introducing songs.
Let's play a clip on how that sounds.
That was the brand new single floating by Schoolboy Q featuring 21 savage.
We'll be rolling all of these different types of expressions into Alexa kind of domain by domain over the coming months and years.
The other thing we've found around voices over the past year is the customers are incredibly enamored with celebrity voices.
Whether that's Kelsey [UNKNOWN] Chelsea Candler doing morning greetings on behalf of Alexa or Ed Sheeran playing a duet with Alexa, which is super fun, and we love doing these but up till now we've actually pre-record the recordings from the celebrities themselves.
Neural text to speech allows us to [UNKNOWN] Celebrities voice with their permission of course, and we wanted to see how this would go.
So we picked the most iconic voice we could to use for the first celebrity voice.
And that has to be Samuel L Jackson.
Sam can answer your voice with jokes.
Lets you know it's raining, his personality, timers, alarms, play music and a lot more when you add this.
And it's in his own personality.
So I will warn you, there's an explicit version of this and a non explicit, I will leave it up to the audience to choose which one you want to hear.
And we shot a little bit of a video on how this came to be, so why don't we roll that real quick.
Just like we practiced.
I'm always ready.
Today in Los Angeles, it's 85 degrees.
Say my name, Woo hoo.
Aw, an't we organized?
[LAUGH] I'm not reminding you.
Not bad for a rookie.
Say rookie again.
I dare you.