Apple Watch can now replay what someone just said to you. The feature entered developer beta testing on November 2. It forms part of Apple’s new Audio Intelligence suite for watchOS. Four AI-powered features are rolling out for Apple Watch Series 12 wearers.
The headline feature is called Live Rewind. It lets users double-press the digital crown to see the last 15 seconds. The watch then displays what was said as a text snippet. This works during any in-person conversation the wearer participates in.
The snippet disappears permanently after 30 seconds if ignored. However, users can save it to the Siri app for later. They can also tap “Ask Siri” to query something mentioned. For instance, they could ask about an unfamiliar term a waiter mentioned. The second major feature pushes ambient listening further. Called Siri Recaps, it takes high-level notes of conversations automatically. Importantly, Apple says it does not identify individual speakers. It also avoids word-for-word transcriptions entirely.
Instead, Siri Recaps summarizes conversations into short bullet points. These capture key facts like dates, deadlines, and pricing information. Users review saved recaps at the end of each day. Anything not saved deletes itself within seven days automatically.
Apple built notable privacy safeguards into the architecture. Raw audio never leaves the secure enclave on the watch. A dedicated hardware-protected memory space processes it instead. The paired iPhone then converts audio into condensed text locally. Thankfully, only the condensed text reaches Apple’s Private Cloud Compute for summarization. That summary and all interim data delete from the cloud immediately afterward. Consequently, Apple says it never accesses the raw audio itself.
Additional privacy signals alert people nearby too. A chime sounds when Live Rewind activates on the watch. An animation also displays on screen simultaneously. Together, these signal that something was just recorded.
The suite also includes sound recognition for deaf users. A separate feature uses Shazam to identify songs playing nearby. Both features require the S11 chip in Series 12 or Ultra 4.
The beta rolls out to developers first in iOS 27.3. A public beta follows before the wider consumer release.
