The key facts
- Audio Intelligence entered developer beta testing in iOS 27.3; it was announced in September at the Surprise and Shine event
- Live Rewind shows the text of the last 15 seconds of a conversation; Siri Recaps creates brief summaries without recording raw audio
- The features run on Apple Watch Series 12 and Ultra 4 with the S11 chip and require an iPhone with Apple Intelligence
What has started being tested
The Audio Intelligence suite of AI features for Apple Watch, first shown in September at the Surprise and Shine hardware event, has entered beta testing — initially for developers in the iOS 27.3 update, TechCrunch reports. The suite includes four optional features for Apple Watch Series 12 owners: recognition of important sounds (sirens, alarms, doorbells — primarily for deaf and hard-of-hearing users), identification of playing music via Shazam, and two conversation-related features — Live Rewind and Siri Recaps. The features will later appear in the public beta and then become available to everyone.
Live Rewind and Siri Recaps
Live Rewind lets you see a text snippet of the last 15 seconds of what was said with a double press of the Digital Crown — for example, spoken directions or a list of dishes. The snippet either disappears forever after 30 seconds or is saved to the separate Siri app that appeared with iOS 27; below the snippet there is an 'Ask Siri' button for follow-up questions. When pressed, a sound plays and an animation appears — so those around understand that what was said has been captured. TechCrunch notes that these are precisely the features that look most controversial against the backdrop of criticism of recording devices like Meta's AI glasses, and Apple is trying to make them useful without infringing on the privacy of those nearby.
Siri Recaps
Siri Recaps goes further: it creates brief summaries of conversations — a short sentence and bullets with key facts (dates, deadlines, prices). Apple emphasizes that the feature does not identify speakers, does not store raw audio recordings, and does not produce a verbatim transcript; in style it imitates the user's own notes, without emotions or tone. The summaries are suggested for use at work meetings, parent-teacher conferences, or workouts. At the end of the day, the system offers to review what was recorded and choose what to keep — in the Notes app or, if the notes are personal, in Journal; anything not saved is deleted within seven days. The feature can be turned on manually or on a schedule (for example, only at work), and disabled at any time via Control Center; it does not work while listening to music on headphones with noise cancellation. According to Apple, the summaries deliberately skip sensitive data: government identifiers, financial and authentication information, and statements about self-harm.
How data processing works
The features run on Apple Watch Series 12 and Ultra 4 and rely on the S11 chip with a built-in microphone, Apple Intelligence models on the iPhone, and Private Cloud Compute. Lightweight models on the S11 determine that a conversation involving the owner is taking place; the raw audio buffer is processed in the secure enclave — a hardware-protected memory area inaccessible to the rest of the system — and is transmitted only to the paired iPhone via a secure mechanism. On the iPhone, the audio is turned into 'condensed' text without filler words, encrypted, and sent to Private Cloud Compute to create the summary, after which the data is deleted from PCC. The finished summaries are available only to the user, not to Apple, and are encrypted when synced between devices. Apple also states that the features do not affect the battery life of the watch or the phone.




