Key facts
- Live Rewind shows the last 15 seconds of speech as text after a double-press of the digital crown, with a chime and animation.
- Unsaved Live Rewind snippets delete automatically after 30 seconds; unsaved Siri Recaps delete within seven days.
- Apple said Siri Recaps does not identify speakers, save raw audio or produce word-for-word transcriptions.
- Siri Recaps omits government identifiers, financial and authentication data and language promoting self-harm, Apple said.
- The features run on the S11 chip, the iPhone's Apple Intelligence models and Apple's Private Cloud Compute, Apple said.
- Apple said the features should not affect watch or phone battery life and will later reach public beta.
Apple released its Audio Intelligence features for the Apple Watch into beta testing for developers in the iOS 27.3 update, after announcing the technology at its Surprise and Shine hardware event in September. The set covers four opt-in features for Apple Watch Series 12 and Apple Watch Ultra 4 wearers: sound recognition, song recognition, Live Rewind and Siri Recaps.
Sound recognition is designed to help deaf and hard-of-hearing users identify sirens, alarms and doorbells from the watch. Song recognition uses Apple's Shazam app to identify music playing in the background and show the song name and artist.
The two features that listen to conversations are Live Rewind and Siri Recaps. With Live Rewind, a wearer double-presses the digital crown to see the last 15 seconds of speech as a text snippet on the watch face. An audible chime and an on-screen animation accompany the gesture, which Apple said is meant to show people nearby that something was recorded. The snippet disappears permanently if not saved, or can be stored in the standalone Siri app shipped with iOS 27. Unsaved snippets delete automatically after 30 seconds, and an Ask Siri option below the text lets users query something that was said.
Siri Recaps summarizes a conversation as a short sentence and bullet points covering items such as dates, deadlines and pricing. Apple said the feature does not identify speakers, save raw audio or provide word-for-word transcriptions, and that it omits government-assigned identifiers, financial and authentication data, certain personal identifiers and language promoting self-harm. Recaps do not run while the wearer listens to music on headphones or uses noise cancellation. Users are prompted to review saved recaps, which can go to the Notes or Journal apps and delete themselves within seven days if unsaved. Recapping can be enabled manually or on a schedule, such as only during work hours, and can be switched off from Control Center at any time.
Apple said the features rely on the S11 chip, the watch's built-in microphone, Apple Intelligence models running on the paired iPhone and Private Cloud Compute. A dedicated audio buffer in a secure enclave on the watch processes raw audio, which is accessible only to the paired iPhone through a secure pairing mechanism, Apple said. Lightweight models on the S11 chip detect a conversation involving the wearer, and on-device processing on the iPhone converts the audio into condensed text with filler words removed. That text is encrypted and sent to Private Cloud Compute to generate the recap, which is deleted from the service immediately afterward, according to Apple.
Apple said the features should not affect the battery life of the watch or phone, and that they will later move to a public beta before general release.