Sonos will launch a new voice assistant next month, which will coexist in its connected speakers with Alexa and the Google Assistant. Despite being a few years late, and even though it was created by a much smaller company than Amazon and Google, the Assistant still manages to stand out, as far as I could see during a test before launching the service.
A private assistant
The greatest strength of Sonos’ Assistant is its private design. Unlike Amazon, Apple and Google which upload your recordings to the cloud for analysis, all voice recognition here is done on the speaker directly.
The recording is never stored, either locally or on Sonos servers. Only the names of songs, artists, albums, lists or genres are sent to the various compatible music services (like Apple Music, but not Spotify at launch).
The privacy benefit is obvious, but it’s not the only one. Thanks to local processing, recognition can also be done faster, according to Sonos. “The speed is enabled by the fact that the processing is local, so there is no round trip to the cloud. We also process each fragment of sound one after the other, without waiting for the end of the sentence,” explained Alice Coucke, head of machine learning at Sonos.
In practice, the assistant did indeed seem fast to me, but it will be interesting to try it more thoroughly to quantify the improvement.
The private aspect of Sonos’ Assistant offers another advantage: it allows listening for a few more seconds after the initial request, so that the user can make additional requests, without repeating the wake word (Hey Sonos ). “If we processed the data in the cloud, we would record very long portions of conversations. There would be problems with respect to privacy, ”notes Alice Coucke.
Extended listening was not designed for this, but it could also be seen as an accessibility measure. As a stutterer, I often have to start a request again because Siri or the Google Assistant got impatient. An assistant who processes each sound one after the other locally, and which records longer, could be a game-changer for people in my situation (and for children, who also struggle to say the sentences in the required times).
Technically, the Sonos voice assistant would offer another advantage, according to the company, that of being dedicated to music. This could therefore include additional requests for music, without having to worry about all the other requests, which would increase its efficiency. It seems logical, but here too, the assistant will have to be tested for longer to measure this improvement.
That said, I still would have liked the assistant to include a few more important queries, like the weather.
An English voice with a lot of style
Voice is another big difference between Sonos’ Assistant and those from Amazon, Apple and Google. While the techno giants opted for neutral and generally feminine voices, Sonos chose that of actor Giancarlo Esposito (Breaking Bad, The Mandalorian).
His answers have style and are given in an assertive tone. It’s not the first time that an assistant has used a familiar voice (GPS navigators have been doing this for several years), but it’s still amazing.
Do I like it? We’ll see as time goes on. So far, I’m more surprised than amazed.
What is the Sonos voice assistant worth in French?
Sonos voice assistant will launch on 1er June in American English, then in French from France later this year.
Good news, even with my sharp French accent, the assistant had no problem understanding me when I asked him for songs in English. Out of curiosity, I also asked (in English) for songs and artists in French: in a fraction of a second, the assistant understood that something was wrong, and redirected me to the mobile application.
It bodes well. It shows me that when French is added, the assistant should be able to handle bilingual requests.
In other good news, the team responsible for the development of the assistant is that of the former French company Snips, acquired by Sonos in 2019. In other words, those who created and tested the assistant throughout its development also speak with a French accent, and also (I presume) listen to French-speaking music on occasion.
Unfortunately, adapting the assistant to a new geography is a complex process, and it is clear that there will probably be many markets larger than Quebec to target before the assistant is adapted to the language of Dédé Fortin and by Kevin Parent.
It remains to be seen if we will still be able to make ourselves understood by then.