watchOS 27.3 beta turns on Siri Recap and Live Rewind audio features

Apple's watchOS 27.3 beta now lets testers try Siri Recap and Live Rewind, the two headline Audio Intelligence features the company promised when it unveiled the Apple Watch Series 12 and Apple Watch Ultra 4 in September. The release moves one of Apple's most privacy-sensitive product bets from keynote slide to working software. And it does so on hardware that, by Apple's own description, keeps a microphone constantly listening to its surroundings.

That combination is the story. Apple has built an always-on audio pipeline into a wristwatch, and the watchOS 27.3 beta is the first time the public can test how it behaves in daily life.

What do the new features actually do?

Siri Recap generates high-level summaries of conversations. Apple Intelligence produces a title and key points for each conversation, and users can pull those summaries up in the Siri app whenever they need to jog their memory. Think of it as a searchable index of what was said around you, not just to you.

Live Rewind addresses a narrower problem: the moment you missed. The feature converts the last 15 seconds of a conversation into text. Double-press the Digital Crown and the snippet appears on screen. You can ask Siri about the text, or save it to the Siri app for later.

Apple's examples are mundane and practical. An airport announcement you couldn't quite hear. A line in a presentation that slipped past you. Instead of asking someone to repeat themselves, you rewind.

The features are exclusive to the Apple Watch Series 12 and Apple Watch Ultra 4. The reason is silicon: the S11 chip inside those models can process audio in isolation from the rest of the device. That isolation underpins Apple's entire privacy argument.

How does Apple defend an always-on microphone?

The microphones on the Series 12 and Ultra 4 constantly pick up audio from their surroundings. That hardware reality also supports other Audio Intelligence capabilities Apple has shipped or announced, including:

• Sound recognition for events such as alarms and a crying infant
• Faster Shazam song detection
• Live Rewind's 15-second audio-to-text conversion
• Siri Recap's conversation summaries

An always-on microphone on a wearable invites obvious questions. Apple's answer is architectural. The company says these features don't record or store audio. The S11 chip processes the audio in isolation, then deletes it right away. Apple claims its operating systems can't access the raw audio — and neither can apps, users, or Apple itself.

That last clause carries weight. Apple is asserting that even it cannot retrieve what the microphone hears. If that claim survives independent scrutiny, it would separate the Watch from smart speakers and phones that upload or retain audio for processing. If it doesn't, the feature set becomes a liability.

What guardrails ship with the features?

Apple has layered controls and signals around the audio pipeline. None of the features run by default in a way users can't see or switch off. According to the company:

• Users can opt out of each Audio Intelligence feature individually
• Some features must be manually enabled before they work at all
• Audio Intelligence doesn't identify or attribute speakers in a conversation
• Live Rewind plays a chime when in use, even if the Watch is set to silent
• The display shows an animation and a microphone indicator during use
• Siri Recap summaries won't surface sensitive information, such as financial details

The chime detail is notable. Apple chose to make Live Rewind audible even in silent mode, which means bystanders get at least one signal that the feature is active. The speaker-anonymity limitation similarly narrows what the summaries can become: a record of what was said, but not of who said it.

The financial-details exclusion in Siri Recap suggests Apple has already mapped the categories where a wrong summary could do real damage. A garbled recollection of a dinner plan is an inconvenience. A garbled recollection of a bank transaction is something else entirely.

Who can actually try the beta?

The eligibility list is short and specific. Testers need:

• An Apple Watch Series 12 or Apple Watch Ultra 4
• A compatible iPhone with Apple Intelligence enabled — meaning an iPhone 16 or later, but not the iPhone 16e
• Willingness to install a beta build of watchOS

The geographic and language restrictions matter as much as the hardware ones. Siri Recap and Live Rewind will initially be available only in English, and not in the EU. Both features are also subject to daily usage limits.

The EU exclusion follows a familiar pattern for Apple Intelligence capabilities, which have arrived in Europe later than elsewhere amid regulatory friction. The daily usage caps, meanwhile, signal that Apple is treating the rollout cautiously — throttling demand while it watches how the on-device processing performs at scale.

Why this beta matters beyond the Watch

Apple is staking a specific claim here: that useful ambient audio AI and strict privacy guarantees can coexist on a consumer device, without audio ever leaving an isolated chip. The watchOS 27.3 beta is the first broad public test of that claim, and it arrives as regulators, competitors and researchers all wrestle with where always-listening assistants should draw lines.

The stakes run past the wrist. If Siri Recap and Live Rewind ship cleanly — with no leaks, no bypasses, no privacy incidents — Apple gains a defensible template for audio AI it can extend across its ecosystem. If the guarantees wobble, the company hands ammunition to critics of ambient computing at exactly the moment it's pushing hardest into it. What testers find in this beta, and what Apple reports about the S11 chip's isolation in practice, will shape how confidently the features move from English-only beta to a full public release.