Chevron icon It indicates an expandable section or menu, or sometimes previous / next navigation options. HOMEPAGE

Apple has revealed how it makes its AI assistant Siri's voice sound how it does

olivia palermo actress
Deciding how Siri sounds is no easy task. D Dipasupil/Getty Images for Advertising Week New York

If you've ever used an iPhone, you're almost certainly familiar with Siri, Apple's virtual assistant.

Advertisement

It can answer question, fulfill tasks, and manage calendars — most commonly with a soothing female voice.

But how does Apple make Siri sound like it does?

Thanks to some new research papers published by the Californian technology company, we now have a better idea. (We first saw them via The Register.)

Most commonly using a soothing female voice, Siri answers questions, fulfills tasks and manages users' calendars. But it comes in a number of voice options — male or female, with accents including American, British, and Australian — and these are based on human voice actors.

Advertisement

In selecting these actors, "first and foremost, a voice must be perceived as being compatible with the Siri personality," Apple engineers wrote.

They don't elaborate on exactly what the "Siri personality" is — but it typically comes across as restrained, neutral and professional, with the occasionally dryly delivered joke for those who know what to ask.

Once a suitable voice talent is found, between 10 and 20 hours is recorded of their voice. "The recording scripts vary from audio books to navigation instructions, and from prompted answers to witty jokes," Apple's Siri team wrote in a blog post.

"Typically, this natural speech cannot be used as it is recorded because it is impossible to record all possible utterances the assistant may speak." As such, it's then chopped up into constituent blocks that can be put together using to generate new speech — even words that the actors never uttered.

Advertisement

The tricky part is constructing Siri's speech in a way that sounds natural and "human" — and to do this, Apple uses a number of artificial intelligence (AI) techniques. The company's researchers go into more technical detail on how they achieve this in newly published papers.

These developments seem to be paying off: In tests, Apple wrote, "the new voices were rated clearly better in comparison to the old ones." You can can hear examples of how Siri has evolved from iOS 9 to iOS 11 at the bottom of the page.

Apple has historically been highly secretive, rarely talking about its inner workings. but in December 2016, it announced it would allow its artificial intelligence researchers to start publishing their work publicly and engage more in the broader academic community, as it tries to attract more AI experts to join the company.

Here's the full paper on Siri:

Apple Siri Artificial Intelligence
Advertisement
Close icon Two crossed lines that form an 'X'. It indicates a way to close an interaction, or dismiss a notification.

Jump to

  1. Main content
  2. Search
  3. Account