Apple starts beta testing Watch conversation rewind and recap features

Apple’s first developer beta versions of iOS 27.3 and watchOS 27.3 include Live Rewind and Siri Recap.[1][2] Apple’s Audio Intelligence suite also includes Sound Recognition and Music Recognition with Shazam.[3] With Live Rewind, a wearer double-presses the Digital Crown to display a text snippet of the previous 15 seconds; they can ask Siri about the snippet or save it to the Siri app.[3] Using the feature plays an audible chime.[2]

Siri Recap listens to conversations during the day and creates high-level summaries intended to help users remember what was said.[3] Apple says the feature does not provide word-for-word transcripts or identify speakers.[4] Users can schedule a recap by time, location, or both.[2] Summaries can be automatically deleted after up to seven days unless users explicitly save them.[3]

Live Rewind and Siri Recap require an iPhone 16 or later running Apple Intelligence and Siri AI.[2] The features initially support English only and are unavailable to users younger than 13.[2] Apple support material identifies Apple Watch Series 12 and Apple Watch Ultra 4 in connection with Audio Intelligence features.[3] Apple planned a public beta for broader testing before making the features publicly available; the full watchOS 27.3 rollout was expected in December 2026.[4][2] Apple had previewed Audio Intelligence at a September product event before beta testing began.[4]

Apple says every Audio Intelligence feature is opt-in.[3] Apple also says audio recordings are not created or stored, and that apps, the operating system and Apple cannot access audio.[2] Siri Recap summaries are created through Apple’s Private Cloud Compute service.[2] Apple says recaps are designed to omit sensitive information such as government identifiers, financial and authentication data, and some personal identifiers.[4] Apple says the features should not affect Apple Watch or iPhone battery life.[4] Teams assessing the tools should weigh their summary function against the lack of transcripts and speaker identification.[3][4]