Apple Watch's conversation rewind and AI recaps enter developer testing, TechCrunch reports
Live Rewind turns the last 15 seconds of speech into text; Siri Recap creates high-level notes. The reported developer beta follows September's announcement and carries device, language and privacy limits.
Miss a name, a direction or the last thing someone said? Apple’s Watch is moving its conversation-memory pitch into testing. TechCrunch reports that Audio Intelligence is rolling out to developers in the iOS 27.3 update, including Live Rewind and Siri Recaps. The new milestone is developer access after September’s announcement, rather than a general release for all Watch owners.
Sarah Perez’s report describes four opt-in features: sound recognition, music identification, short conversation rewind and high-level conversation notes. Apple’s existing English support documentation explains the intended behavior and requirements, but still uses forward-looking beta language. The current developer-rollout claim therefore comes from TechCrunch’s reporting.
A text rewind, plus a different kind of recap
Live Rewind displays a text snippet of speech from the previous 15 seconds when the wearer double-presses the Digital Crown. Users can ask Siri about the snippet or save it. Apple says activation produces an audible chime, a display animation and a microphone indicator, including when the Watch is on silent.
Siri Recap takes a different approach: short summaries intended to resemble notes, rather than a speaker-attributed transcript. Users can enable it manually or set a time or location schedule, and turn it off in Control Center. Unsaved recaps auto-delete after seven days. Apple’s documentation says unsaved Live Rewind snippets disappear approximately 30 seconds after the Watch display dims, unless saved or used with Ask Siri.
Apple also warns that generated snippets and summaries can omit or misunderstand important information. Their presence is not evidence that an agreement, instruction or recollection has been captured accurately.
New hardware, paired-phone processing and privacy claims
Apple lists Watch Series 12 and Ultra 4 for these features, paired with a supported Apple Intelligence iPhone, with English support and Siri AI beta requirements. Its support page lists supported iPhone 16, 17 and 18-era models, including iPhone Air and iPhone Duo. Older iPhones eligible for sound or music recognition are not automatically eligible for conversation rewind and recaps. Apple says Live Rewind and Siri Recap are unavailable under age 13, and Child Accounts under 18 require parental or guardian enablement.
The company describes protected audio buffers in a Secure Exclave, on-device speech processing and Private Cloud Compute for recap generation. Condensed conversation text goes to that cloud service to make the summary. Apple says raw audio is not stored as an accessible recording, and saved text syncs with end-to-end encryption under its documented iCloud conditions.
Those are Apple’s architecture and privacy claims, not an AITrending security audit. They also do not make the feature entirely offline. The support page notes that sensitive-content filtering may miss information and asks users to consider nearby people in private conversations. TechCrunch says public-beta testing will follow before broad availability; the report does not supply a firm public-release date.