Apple Watch Audio Intelligence beta adds AI conversation rewind and recap

Apple is rolling out Audio Intelligence features for Apple Watch Series 12 in iOS 27.3 beta, introducing AI-powered tools for rewinding conversations and generating recaps. The feature set includes Live Rewind (captures the last 15 seconds of speech as text) and Siri Recaps (creates bullet-point summaries of conversations without word-for-word transcription). Apple emphasizes privacy protections through on-device processing, secure enclaves, and automatic deletion of audio data.

Key points

• Live Rewind lets users double-press the digital crown to capture the last 15 seconds of conversation as a text snippet, with an audible chime and visual animation to signal recording
• Siri Recaps creates high-level summaries of conversations with bullet points covering key information like dates, deadlines, and pricing—without identifying speakers or saving raw audio
• Audio processing uses the S11 chip's secure enclave, on-device iPhone models, and Apple's Private Cloud Compute, with immediate deletion after processing
• Snippets auto-delete after 30 seconds unless saved; recaps delete after 7 days; users can schedule when recapping runs (e.g., only during work hours)
• Additional features include sound recognition for deaf/hard-of-hearing users and background music identification via Shazam integration

How to use it

• Double-press the Apple Watch digital crown during a conversation to retrieve the last 15 seconds as text—useful for catching directions, restaurant specials, or someone's name you missed
• Enable Siri Recaps on a schedule (e.g., only at work or during specific hours) to automatically generate conversation summaries with key dates, tasks, and factual information from meetings or conferences
• Review saved recaps at the end of each day or when unlocking your watch in the morning, then choose to export them to Notes or Journal apps for long-term reference
• Turn off recapping anytime from Control Center, even when running on a schedule, to respect privacy in sensitive conversations

Has anyone tested ambient AI recording devices for note-taking? How do you balance the convenience against privacy concerns for people around you?