Apple Watch’s AI features for rewinding and recapping conversations begin beta testing

Apple has begun beta testing its AI-powered "Audio Intelligence" features for the Apple Watch, releasing them to developers in the iOS 27.3 update. The capabilities were first unveiled at the company's Surprise and Shine hardware event in September and comprise four opt-in features for Apple Watch Series 12 owners: detecting important sounds, identifying songs, catching up on conversations, and taking notes.

Two of the features address accessibility and ambient media. Sound recognition alerts deaf and hard-of-hearing users to events such as sirens, alarms, and doorbells, while song identification draws on Apple's Shazam app to surface the track title and artist of music playing in the background.

The conversational features carry more scrutiny. Live Rewind lets a wearer double-press the digital crown to see a text snippet of what was said over the previous 15 seconds — useful, Apple suggests, when someone is giving verbal directions or reading off a restaurant's specials. The snippet can be saved to the standalone Siri app that shipped with iOS 27, or dismissed, in which case it is deleted automatically after 30 seconds. An "Ask Siri" option below the text lets the wearer ask follow-up questions about an unfamiliar term or option.

Apple has also built in signals intended to make the recording visible to bystanders: double-pressing the crown triggers an audible chime and an on-screen animation, a combination of gesture, sound, and display meant to indicate that something said has been captured. This design choice comes as AI recording devices, including Meta's AI glasses and other AI wearables, have drawn public backlash.

Siri Recaps goes further by generating high-level notes from a conversation, delivering a short summary sentence plus bullet points that pick out key facts such as dates, deadlines, pricing, or payment plan details. Apple stresses that the feature does not identify speakers, does not retain raw audio, and does not produce word-for-word transcriptions — a boundary the company is drawing against fuller transcription tools on the market.

What remains open as the beta progresses is whether these consent cues will satisfy people nearby, and how developers and users will respond to a summarization layer that deliberately omits speaker attribution and verbatim text. Adoption in the developer beta of iOS 27.3 will indicate how far Apple pushes the feature before a wider release, and whether the 30-second auto-delete window and opt-in framing hold up under real-world use.

keyFacts:[