THE ESSENTIALS IN 30 SECONDS

  • Presets, rolling out in Sonos app 91.0 from November 3, save a combination of rooms, music source and volume that can be recalled with one tap.
  • The 27voice early-access waitlist opens November 6 for US English users; Gizmodo reports access begins November 12, and approval is not guaranteed.
  • Unlike Classic Sonos Voice Control, which runs on the speaker, 27voice sends requests to Sonos's cloud and uses some third-party AI models.

Sonos is rolling out two pieces of its Sonos 27 software this month: Presets, a way to save and recall listening setups, and early access to 27voice, an AI voice assistant that replaces on-device processing with a cloud language model for users who opt in.

Presets and other app changes

According to Tech Times, Presets arrived with version 91.0 of the Sonos app, which began rolling out on November 3 and may take days or weeks to reach every account. A preset stores which rooms are playing, the music source and the volume, and restores them with one tap from the home screen. Each account can keep up to 50 presets, shared across signed-in devices in the household. For now, they can only be triggered from the mobile app, not by voice or from the web app, and they do not capture EQ settings. They require S2 hardware able to run Sonos 27. The same update adds an iOS-only System Health screen showing network and connection details.

How 27voice works

Gizmodo reports that the 27voice waitlist opens on November 6 and that access starts on November 12; Engadget describes November 6 as the start of early access, with users requesting entry in the app and waiting for approval. Early access is limited to US households and US English, and Sonos staff say not everyone who signs up will be selected. A full release with more languages is targeted for 2027.

The assistant accepts natural-language requests, such as asking for music by mood or genre in a given room, and can group speakers, move audio between rooms, set timers, control supported smart-home devices and answer general questions. Sonos says it costs nothing extra and needs no subscription.

The privacy trade-off

The main difference from the existing assistant concerns where speech is processed. Classic Sonos Voice Control, which stays available, handles wake-word detection, speech recognition and understanding entirely on the speaker. With 27voice, only the "Hey Sonos" wake word is detected locally; the request then goes to Sonos's cloud, where a Sonos-tuned language model processes it, and some open-ended requests are passed to third-party AI services that Sonos does not name. The company says recordings and transcripts are deleted once a request is fulfilled. 27voice requires an internet connection and does not work on portable speakers in Bluetooth mode, and a household can run only one of the two assistants at a time.