Apple’s Audio Intelligence features for Apple Watch Series 12 and Apple Watch Ultra 4 began developer beta testing in iOS 27.3 on 2 November, TechCrunch reported. Announced at Apple’s Surprise and Shine hardware event in September, the opt-in features let wearers retrieve words from a conversation they have just heard or receive a short account of one. The beta puts those listening features into developers’ hands for testing.
Key points
- Live Rewind displays words from the preceding 15 seconds after a double-press of the Digital Crown.
- Siri Recaps makes brief conversation summaries rather than word-for-word transcripts.
- Apple says processing runs across the watch, paired iPhone and Private Cloud Compute.
- Unsaved snippets disappear after 30 seconds; unsaved recaps are deleted within seven days.
Live Rewind retrieves the last 15 seconds
Live Rewind starts with a double-press of the watch’s Digital Crown. It displays a text snippet of what was said in the previous 15 seconds, allowing the wearer to check words they missed. The action also produces a chime and an animation on the watch, which Apple intends as a signal to other people that speech has been captured. A snippet disappears after 30 seconds unless the wearer saves it to the Siri app.
Remembering spoken directions could involve checking the last 15 seconds in writing rather than asking for them again. If the words were worth keeping, saving the snippet would make it available later; leaving it alone would mean it disappeared after 30 seconds.
Apple says an Ask Siri option beneath the snippet will allow questions about something said. Live Rewind also transfers audio to the paired iPhone for processing into the text shown on the watch. The brief window and the need to press the Crown make this a way to revisit a particular moment, rather than a running transcript of everything said nearby.
Siri Recaps condenses conversations into notes
Siri Recaps takes a different approach to the same listening capability. Apple says it produces a short sentence and bullet points about a conversation, potentially identifying dates, deadlines or other factual details. That distinction matters to anyone relying on a recap: the output is a set of selected points, rather than a record from which every exchange could be recovered.
Wearers can turn recapping on manually or set a schedule, and can switch it off through Control Center while a schedule is active. TechCrunch reported that recaps are not made while headphones are playing music or using noise cancellation. When the feature is enabled, prompts invite the wearer to review recaps at the end of the day and on first unlocking the watch in the morning. Notes can be saved to Notes or exported to Journal; otherwise, they are deleted within seven days.
Apple says Siri Recaps is designed to leave out government-assigned identifiers, financial and authentication data, certain personal identifiers and language promoting self-harm. Those exclusions sit alongside a broader expansion of Siri’s capabilities: AI Affairs previously reported its addition of five languages in Apple’s 27.2 updates. For companies building services around a recap, the described output offers selected facts rather than a transcript or an identified account of who said them.
The S11 chip starts audio processing
Audio Intelligence uses the S11 chip and microphone in the two watch models, models on the paired iPhone and Apple’s Private Cloud Compute service. Lightweight models on the watch detect when the wearer is part of a conversation. A dedicated audio buffer processes raw sound inside the chip’s secure enclave — a hardware-protected space separated from the rest of the operating system. The audio then reaches the paired iPhone through a secure connection between the devices.
On the iPhone, processing turns speech into condensed text by removing filler words. That text is encrypted before being sent to Private Cloud Compute, which creates the recap and returns it to the phone. Apple says the service deletes the recap and the intermediate summaries used to make it immediately after processing. The company also says recaps on the iPhone are accessible only to the wearer and are encrypted when synced between their Apple devices.
This division of work gives the watch the job of detecting and buffering speech, while the paired phone prepares text for cloud summarisation. It also makes the paired iPhone part of how the features operate. Apple says the listening features should have no effect on watch or phone battery life. The company expects developer testing to be followed by a broader public beta before general availability.
