Personal Voice is an accessibility feature introduced with iOS 17 in September 2023 that lets users create a synthetic voice that sounds like their own. It is designed primarily for people who are at risk of losing their ability to speak due to conditions like ALS. After recording 150 short phrases (roughly 15 minutes of audio), the device processes the recordings overnight using on-device machine learning and creates a voice clone the user can speak through in calls, FaceTime, and apps.

The technical pipeline involves fine-tuning both a neural text-to-speech acoustic model and a vocoder model (software that generates the actual audio output) on the device overnight while charging. Apple's machine learning researchers published the technical details of this system, noting that on-device training — rather than cloud processing — is central to preserving privacy. The finished voice is encrypted and stored locally, only accessible via Face ID, Touch ID, or device passcode. Users can optionally sync their Personal Voice across devices via iCloud with end-to-end encryption. Third-party assistive communication apps can request access to Personal Voice with the user's permission.