How Sonos Made Alexa Sound Amazing – WIRED
Normally, Amazon’s Alexa voice assistant communicates through a ring of blue lights. They blink to life when Alexa hears its name, then whirl around the top of an Echo device while Alexa thinks. But when Sonos started working on integrating Alexa into its multi-room wireless music system a year and a half ago, though, that setup posed a problem. Most Sonos users have more than one speaker, and a surround sound system full of blinking lights would turn their homes into an ersatz light show. Plus, it felt wrong: Sonos is an audio company, betting its future on sound as the interface for all the tech in your home. There had to be an audible way for Alexa to let you know it’s listening.
The team seized on a setting buried in the Alexa app, which enables what Amazon calls “sound cues.” Rather than lighting up, Alexa chirps out a bright, rolling tone, like you’ve just found an item in a video game. It’s Alexa’s way of saying hello.
Even this didn’t work for Sonos, though. Amazon created its sound cues for the Echo and Dot, which you might generously define as “meh” speakers. Sonos wanted Alexa in its high-end system, and in a new speaker built specifically with Alexa in mind. “So we [told Amazon] hey, we’re making a speaker that has pretty broad fidelity, when compared to something like the Dot,” says Dayn Wilberding, the UX creative director at Sonos. “And so the wake sound you’re going to play…know that it’s going to be played on this big system.” Amazon’s response: no problem. The Alexa team actually had a better-sounding version ready to go. They’d just never needed it before.
That tone is the first thing you’ll hear when you talk to the latest Sonos device, the Sonos One. The $199 speaker is its first to directly support Alexa, and the first-ever Sonos product with a microphone inside, a long-overdue upgrade to the company’s hardware. It’s also the beginning of a new Sonos, more than 15 years after the company began developing the multi-room wireless system that won it millions of fans. Sonos believes that as more devices come online and voice interfaces improve, your home’s sound will be just as important as its heating system or curb appeal. “We talk a lot about being the speaker for anything that makes sound inside your home,” says Antoine Leblond, the company’s vice president of software. “That’s the thing we’re really moving toward.”
Alongside the new speaker, Sonos also announced today that by 2018 it will support both Google Assistant and AirPlay 2, allowing Sonos users to query Google, use Siri to control their speakers, and to stream tracks from their iPhones. Sonos wants to support all voice assistants, all smart-home gadgets, and everything else that makes sound, just as it tries to support all music services.
But save all the visions of the future for another day. The road is long, requires years of work from companies larger than Sonos, and won’t matter much to Sonos users anytime soon. Not to mention, the first test for Sonos and Amazon was hard enough: making Alexa amazing for music.
Play Me Something Good
Since Amazon launched the first Echo in 2014, people have used Alexa to play music. One 2017 study found music to be the most popular use of devices like the Amazon Echo or Google Home. (Another poll listed music second, only after setting timers.) “Music is integral to Alexa,” says Rohit Prasad, Amazon’s head scientist for Alexa. In a way, it exemplifies everything great about voice interfaces. You just ask for something, and Alexa finds it for you.
The voice-music connection was not lost on Sonos. In fact, it was becoming a problem. Even Sonos users—who love high-fi sound at least enough to spend serious coin on it—were starting to listen to music on smart speakers instead because they were so much easier to use. In 2014, Sonos had accounted for half the Wi-Fi speaker market, but Alexa quickly changed that; nobody sold more connected speakers in 2016 than Amazon. With Apple’s HomePod coming out at the end of 2017, and new Echo products coming all the time, Sonos needed a way in.
Sonos could have built an Alexa skill to let you control volume, change songs, and hear everything through your Sonos speakers. You could even do that with a single cable: out of the Dot, into your Playbase, and ta-da! Voice-enabled Sonos setup.
But the company had bigger ambitions. It wanted to do away with the specific syntax and grammar associated with Amazon skills, so you wouldn’t have to say “Alexa, turn up the volume on Sonos” or “play AC/DC on Sonos.” It wanted to play music anywhere in your house, the way Sonos always has, and let users control their music from anywhere. It wanted the whole thing to feel sensible and natural and easy.
Unfortunately, music turns out to be a wildly complicated voice-control problem. You see P!nk and read “Pink,” and understand “Sk8er Boi” as a phrase that makes sense, but a computer can’t do that. If you want to play Sting, Alexa (or any other service) has to figure out which version of which song on which album on which music app you’re looking for. Sometimes you don’t know what a song is called, just that it’s the one from Moana everyone’s singing. “You have to choose, as Alexa, the best thing the customers want even when they’re not very clear,” says Prasad, Alexa’s chief scientist. “Music humbles you, from a science perspective.”
And so while Sonos improved its speakers, Amazon worked on making Alexa an even better DJ. It allowed for multi-room setups, so you can quickly move music around your house. Thanks to a feature called arbitration, which polls all the devices around you to determine which is the closest and best-suited to respond, you can even say “Alexa, play some music” and it will play wherever you are. This is good for Sonos, good for Alexa (Amazon recently rolled out its own multi-room audio system for owners of multiple Echos), and good for all other third-party devices.
The team is working toward a more natural way for users to interact with their assistant in general. The “Alexa, let me talk to…” grammar helped Amazon scale its third-party capabilities quickly, but hardly qualifies as a natural interaction. When you ask for how to say banana in Spanish, you shouldn’t have to remember the name of a skill, or care which one you’re using. You should just ask.
The One That Goes Doo Be Doo Do
A few weeks ago, I drove down the California coast to see the new Sonos One in action. A team of Sonos executives ushered me into one of the company’s many listening rooms, a small, windowless space set up like the den from a West Elm ad. Inside, there were a preposterous number of Sonos speakers: a Playbase underneath a large TV, a Sub subwoofer against one wall, a surround-sound setup of various Play 1s and 3s. Alone on a coffee table in front of me sat a Sonos One in all black. Two more, in white, flanked the TV.
“Alexa, play ‘Human’ by Rag’n’Bone Man,” Wilberding said. A white light on top of the speaker blinked a few times, and a moment later the song blared from the speaker. Then he asked Sonos to play music on a different set of speakers in the room. A moment later, he reached over, touched the button on top of the One, and the music stopped. Most people, Sonos figures, will have just one or two mic-enabled devices on their network: maybe a Sonos One and an Echo Dot. As long as something can hear you, you can do anything.
“We talk a lot about being the speaker for anything that makes sound inside your home.” — Antoine Leblond, VP of software at Sonos
Wilberding’s demonstration did more than just show off how good the Sonos One sounds. (Very good, for the record, especially in a stereo pair.) It also showed a piece of Sonos’s vision for music, what it calls “continuity of control.” You should be able to control your music any way you want, Sonos believes, at any given time. “Maybe you add another track from your Sonos app,” Leblond says, “Maybe you go into your home automation panel that’s integrated with Sonos, and you hit next. Then you can go to the Spotify app, which also uses that platform, and go to another track. Then you can go to Alexa, and say, what’s playing?” All that better work. And it’s way harder than it looks.
Sonos’ goal was to make sure Alexa on your speakers matches Alexa everywhere else. But simply putting the two sides together took huge amounts of engineering. “Usually, we’ll have someone build a speaker, integrate it with Alexa, and it’s really an Alexa speaker sold by someone else,” says Steve Rabuchin, a VP on the Alexa team at Amazon. “It’s pretty straightforward what we need to do with them.” Sonos, on the other hand, had a fully built cloud service of its own, including hardware, software, and completely different integrations with the same music partners as Alexa. And it wanted to build a platform that would support other assistants like Siri and Google Assistant, rather than be Alexa-only forever.
All of the front-end work came in the form of those tiny details the company prides itself in obsessing over. Like the light underneath the microphone icon, on the top of the Sonos One. The team tested blinking lights, colored lights, lights with icons; some argued it should be a mute button, while others saw it as a power switch. Eventually they found the simplest answer: When the light’s on, the assistant’s on. “The LED is actually hard-wired to the microphones,” says Scott Fink, a Sonos product manager. “So if that light is on, the microphones are on, and if that light is off you know the microphones are off. They literally share the same circuit.”
Even as it launches the Sonos One, a few pieces remain unfinished. You can’t use your voice to play songs through Spotify yet, despite the fact that both Sonos and Alexa support Spotify individually. (All sides say it’s coming in a matter of a few weeks.) Any service Sonos supports but Alexa doesn’t—which is most of them—won’t be voice-controllable yet either. One feature that struck me as obvious, the ability to tell Alexa to play your current song somewhere else, doesn’t work.
But Sonos says this is only the beginning. Microphones add an entirely new dimension to the speaker system, and virtually every Sonos device ever made will support the software integration that lets you control your Sonos system with any Alexa-powered device. And the company says it’s committed to supporting more voice services on the Sonos One, and integrating new products as they come.
“The whole idea of the internet out loud is so much richer than just music, or podcasts, or things like that,” Leblond says. The internet once existed inside your computer. Then it was in your phone. With virtual assistants, it could be everywhere. No one knows quite what that looks like yet, but with a little help from Sonos, it’ll sound just right.