Sonos Keeps Alexa, Google Assistant as It Launches Sonos Voice Control

Nearly three years after it bought Paris-based voice technology company Snips for $37.5 million (see 1911210045), Sonos announced a proprietary voice experience for all mic-equipped Sonos speakers. CEO Patrick Spence underscored privacy in announcing the Sonos-dedicated voice feature on a Wednesday earnings call, saying voice commands are done entirely on the Sonos speaker or sound bar, with no audio or conversation transcript “sent to the cloud, stored, listened to or read by anyone.”

That’s in contrast to the Amazon Alexa and Google Assistant voice engines that Sonos includes on its mic-equipped speakers, which do send data to the cloud. Alexa and Google Assistant will continue to be available on Sonos voice-capable speakers, selectable by the appropriate trigger word. “We designed our experiences to live alongside one another,” Jeff Derderian, Sonos vice president-product program leadership, told us at an embargoed Tuesday pre-briefing in New York. The Amazon and Google options are “great general-purpose voice assistants," Derderian said. "With Sonos Voice Control, we’re not trying to replace that: It’s an and, not an or.”

Sonos Voice Control will be available in the U.S. via a software update June 1 and in France later this year. Initially, it will work with Sonos Radio, Apple Music, Amazon Music, Deezer and Pandora, with more services and markets to follow. “We want to make sure we can get to all of the markets we serve,” Derderian said, saying the experiences have to be “just right.”

Customer feedback shows many smart speaker users choose not to activate voice on their Sonos speakers due to privacy concerns over Alexa and Google Assistant sending information to the cloud, Derderian said. “We know that there are people who want to get the benefit [of voice control] but are concerned,” he said. Finding the “right balance” was part of the genesis of Sonos Voice, he said.

Sebastien Maury, senior director-voice experience, said “your voice, what you are saying, is interpreted on the device,” and no one else hears it. “What you say at home, stays at home.” There’s no continuous data collection as with traditional smart speakers, he said. All processing is done on the speaker.

Derderian said Sonos Voice extends beyond voice commands to “end states you want to get to.” There are “multiple ways to get there because we want the communication to be natural,” he said. In addition to standard track, volume, play and skip commands, Sonos Voice will be able to engage Night Mode and Speech Enhancement on a Sonos sound bar and group and ungroup speakers throughout the home. “The only thing you can’t do is queue up the music,” he said, which has to be done on a smartphone. “You can’t ask for any artists because … you’re working within the confines of the phone," he said. "You can get something going with your phone and then control it from there.”

On how users will know what commands to use, Derderian said, “If we did our jobs right, it should be natural. You should be able to talk the way you expect to talk” and get the desired result. On whether Sonos might partner with a TV or set-top box maker in the future to integrate voice functionality in a home theater setting, Derderian said Sonos is focused on its ecosystem for now, but in the future that would depend on the capabilities of other companies’ hardware. “We planned very carefully the computing power, the memory, RAM, flash -- everything we put under the hood to optimize for our system.”

Several stages occur during the voice control process, Derderian said. “We want a low-power state when we’re listening because we don’t want to burn a lot of energy.” After the wake word triggers the speaker to go on, “We then fire up the rest of the engine to figure out what you said” and whether it’s something the speaker should pay attention to, he said. Sonos’ battery-powered portable Roam and Move speakers are also Sonos Voice compatible.

Part of the reason it took over two years to get to market with Sonos Voice was the amount of choice the company is offering users, Derderian said. “All the different permutations and all the different colloquialisms around how to get to that end state.” The company will continue to improve on the technology and provide software updates, he said.

Sonos is using actor Giancarlo Esposito's voice as the sole option at launch. Spence left open the possibility that the company could monetize the capability in the future by giving customers a choice of voices to purchase. When a Sonos staffer asked a Sonos speaker during Tuesday's demonstration what was playing, Esposito’s voice responded, “Sorry, I don’t know what’s playing right now,” maybe a drawback of not being able to communicate with the cloud.

Sonos also announced a lower priced sound bar Wednesday. The $279 Ray, due June 7, is designed to widen Sonos’ home theater customer base to more budget-oriented users, the company said. It has a dialogue enhancement feature and night mode to minimize loud effects. The Gen 2 Beam ($449) and $899 Arc remain in the line.

Sonos also filled out its $179 Roam portable speaker line, adding “olive,” “wave” and “sunset” colors to join the black and white models. The new colors were due to ship globally Wednesday. And it launched Solarsystym, a channel on Sonos Radio with content curated by singer/songwriter Lorde.