Products & Tools

Apple Watch AI conversation features enter developer beta

Apple's Live Rewind and Siri Recaps enter developer beta in iOS 27.3, with 15-second snippets, seven-day retention, and secure-enclave processing on the S11 chip.

Apple Watch’s AI features for rewinding and recapping conversations begin beta testing
Apple Watch’s AI features for rewinding and recapping conversations begin beta testingLuke Chesser / unsplash
By Marcus Bennett6 min read

Why it matters

  • Apple began developer beta testing of Audio Intelligence features in the iOS 27.3 update, announced at the Surprise and Shine event in September.
  • Live Rewind captures the last 15 seconds of conversation via a double-press of the digital crown; unsaved snippets delete after 30 seconds.
  • Siri Recaps generates conversation summaries without speaker identification, raw audio storage, or verbatim transcripts; recaps delete within seven days if unsaved.
  • The features run on Apple Watch Series 12 and Ultra 4, using the S11 chip, Apple Intelligence models on iPhone, and Private Cloud Compute, which deletes recaps immediately after processing.
  • Features will move to public beta before a general release; Apple says battery life should not be affected.

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

The rollout matters because it places Apple directly into one of the most contested corners of consumer AI: ambient listening on wearables. Devices such as Meta's AI glasses have drawn sustained backlash over recording concerns, and Apple has built its implementation around explicit signals, short retention windows, and hardware-protected processing to avoid the same fate.

What do the four features do?

The sound recognition feature targets deaf and hard-of-hearing users. It identifies important sounds — sirens, alarms, or doorbells — from the watch and surfaces them to the wearer.

The song recognition feature taps into Apple's Shazam app. It detects music playing in the background and identifies the song name and artist.

The remaining two features, Live Rewind and Siri Recaps, are the more potentially controversial ones because they listen to human conversations.

How does Live Rewind work?

Live Rewind lets a user go back in time to see what someone just said, in case the wearer missed it. Apple's examples are practical: someone giving verbal directions, or a server reading a long list of specials at a restaurant.

The mechanics are specific. A user double-presses the Watch's digital crown to see what was said in the last 15 seconds, displayed as a text snippet on the watch face. The user can view the snippet and do nothing, in which case it disappears permanently. Or the user can save it to the standalone Siri app that shipped with iOS 27 for later access — useful for remembering someone's name or taking a quick note from a fast-paced presentation.

Apple says users will also be able to tap an "Ask Siri" option that appears below the text snippet if they have questions about something that was said — like a technical term they do not understand, or perhaps some sauce option the waiter mentioned that they are unfamiliar with.

The privacy signaling is deliberate. When the user double-presses the digital crown, an audible chime sounds and an animation displays on the watch. Apple intends the combination of gesture, sound, and visual display to signal to people nearby that something that was said has been recorded.

The snippets delete themselves automatically after 30 seconds if the user does not save them.

What are Siri Recaps?

Siri Recaps takes the ambient listening technology further. It allows Apple Watch owners to generate high-level notes of their conversations.

Apple states the feature does not identify speakers, does not save raw audio, and does not provide word-for-word transcriptions. Instead, it summarizes a conversation with a short sentence and bullet points identifying key points — depending on the topic, things like important dates or deadlines, pricing or payment plan information, and other factual information.

Apple suggested use cases including work meetings, parent-teacher conferences, and personal training sessions — anywhere a user might need to remember tasks or things that were said. The company frames the feature as mimicking a user's ability to jot down their own notes, without adding emotion or tone.

The feature is aimed at moments when writing notes is impractical, either because the user's hands are full or because the user is concentrating on what is being said. The recaps will not record conversations when the user is listening to music with headphones or using noise cancellation.

Retention and review work on fixed timelines:

  • At the end of the day, the user is prompted to review saved Siri Recaps and choose which, if any, to keep.
  • The same reminder appears when the user first unlocks the Apple Watch in the morning.
  • Saved notes can go to the Notes app, or be exported to the Journal app if they are more personal in nature.
  • If the user does nothing, recaps delete themselves within seven days.

Users can enable Siri Recaps manually or configure a schedule — for example, running recaps only while at work, or ensuring they never run at night. Even on a schedule, the feature can be turned off at any time from the Control Center.

Apple notes that Siri Recaps has been designed to omit sensitive information, including government-assigned identifiers, financial and authentication data, certain personal identifiers, and any language that would promote self-harm.

How does the privacy architecture work?

The Audio Intelligence features for the Apple Watch Series 12 and Apple Watch Ultra 4 rely on a combination of Apple's hardware, software, and services: the S11 chip, the built-in microphone, the Apple Intelligence models that run on the iPhone, and Apple's Private Cloud Compute (PCC).

The always-on processor includes a secure enclave — a hardware-protected memory space that cannot be accessed by the rest of the operating system — where a dedicated audio buffer processes the raw audio. Apple says this allows the audio to be processed without being accessible to anyone, including Apple itself. The audio is only accessible to the paired iPhone through a secure pairing mechanism between the phone and the watch.

The processing pipeline has defined stages:

  1. Lightweight AI models on the S11 chip detect whether a conversation is happening that the wearer is part of.
  2. That detection is transferred to the secure enclave on the iPhone.
  3. On-device processing on the iPhone turns the audio into condensed text — the text of the conversation with the filler words removed.
  4. The condensed text is encrypted and sent to Apple's Private Cloud Compute to create the recap, which is sent back to the phone. The recap and any summarizations used to create it are deleted from PCC immediately afterwards.

The recaps accessible on the iPhone are accessible only to the user, not to Apple, and they are encrypted when synced between multiple Apple devices. Live Rewind follows the same pattern: the Apple Watch transfers audio to the iPhone for processing to create the 15-second snippets.

Why does this matter for the wearables market?

Apple says none of these features should impact the battery life of the watch or the phone — a claim that will face scrutiny as the beta expands.

The features will roll out to the public beta for broader testing before becoming publicly available to all users. For the wearables market, the test ahead is whether Apple's signal-based, short-retention, secure-enclave approach can make conversation-listening AI socially acceptable where rivals' recording devices have failed.

Source: TechCrunch AI

Share this article:

More from Marcus Bennett

Marcus Bennett

Show full bio

Senior reporter covering companies, deals and strategy in artificial intelligence and machine learning at AI In Context.

364 articles

Related articles

  1. Always-On AI Wearables Are Testing the Limits of Privacy Law
  2. Apple's Always-Listening Watch Kicks Off a New Surveillance Fight
  3. Ex-Plaud exec raises $26M for Ambi, an AI that listens all day
  4. Apple ships HomeView, a screen-based smart home hub built around Siri AI
  5. Amazon's Panay Predicts AI Devices Will Kill the App Habit

« Previous article