Apple has begun beta testing Audio Intelligence, a suite of AI-powered features for the Apple Watch that includes Live Rewind, which displays a transcript of the last 15 seconds of a conversation, and Siri Recap, which generates bullet-point summaries of discussions. The first developer beta of watchOS 27.3 is now available for the Apple Watch Series 12 and Apple Watch Ultra 4, with a public release expected in December alongside iOS 27.3.
- Opt-in by design: Audio Intelligence is disabled by default and requires explicit activation through the Watch app on a paired iPhone, along with microphone calibration.
- Privacy-first architecture: Audio is processed in a hardware-isolated Secure Exclave on the S11 chip, and raw audio is never stored or accessible to apps, the operating system, or Apple.
- Bystander safeguards: Live Rewind plays an audible chime and shows a full-display animation even in silent mode, while Siri Recap filters sensitive information and auto-deletes summaries after seven days.
What Audio Intelligence Actually Does
Audio Intelligence comprises four optional features. Sound Recognition alerts users to sirens, alarms, and doorbells, serving as an accessibility tool for deaf and hard-of-hearing users. Shazam Music Recognition identifies songs. Live Rewind, activated by double-pressing the Digital Crown, shows a text snippet of the previous 15 seconds of speech. Siri Recap works ambiently in the background, generating a high-level summary with a title and key points that can be reviewed later in the Siri app.
Live Rewind is designed for moments when you miss something—a waiter reciting specials, a colleague making a quick point—without having to ask for repetition. The snippet can be queried through Siri or saved; otherwise, it disappears after 30 seconds. Siri Recap, meanwhile, can be scheduled by time and location, such as only at the office on weekdays from 1 PM to 4 PM, or toggled manually via Control Center. Unsaved summaries are deleted after seven days.
Engineering Privacy for the People Around You
Apple’s technical approach aims to minimize data exposure. The S11 chip’s Secure Exclave is a dedicated, hardware-isolated compartment that processes audio in complete isolation from the rest of the system and immediately deletes it. Lightweight on-device models detect when a conversation involving the wearer is taking place. For Live Rewind, transcription happens on the paired iPhone; for Siri Recap, summaries are generated through Private Cloud Compute, encrypted end-to-end, and deleted after processing.
“Siri Recap ambiently takes notes on your conversations throughout your day and creates high-level Apple Intelligence–generated summaries that you can use later to jog your memory, so you can stay present in the moment.”
Notably, no Audio Intelligence feature identifies or attributes speakers—summaries will not label who said what. Siri Recap also excludes sensitive information such as financial data or government identifiers. Live Rewind and Sound Recognition are processed on-device unless a snippet is sent to Siri or saved. The features require an iPhone 16 or later with Apple Intelligence and Siri AI, are English-only at launch, and are unavailable to users under 13; Child Accounts under 18 require parental approval.
Hands-On Impressions and Competitive Context
Early testers report that Live Rewind works as described in quiet environments, though the timing of the double-press takes practice. The audible chime is noticeable to bystanders, while the animation is less so. Siri Recap has shown promise in identifying topics and stripping sensitive details, but its summaries can be so sanitized that they lose utility for work purposes. The feature is not intended as a verbatim transcript but as factual notes to jog memory.
Compared with dedicated AI wearables like the Amazon Bee Frontier, Apple’s approach is more conservative. The Bee offers emotional-tone insights and commerce integrations, whereas Apple has emphasized factual recall and privacy. This positions Audio Intelligence as an accessibility and productivity tool rather than an always-on assistant. The deliberate contrast with smart glasses that record without consent is clear: Apple has designed for the person who did not choose to wear the watch.
Ultimately, Audio Intelligence represents a measured entry into ambient computing. The privacy architecture—opt-in controls, on-device processing, audible alerts, and automatic deletion—sets a benchmark for wearable AI. Yet a generated summary remains a summary, with errors and omissions, and users must resist over-trusting it in high-stakes meetings. Apple has built the safeguards; the judgment call remains with the wearer.