How Audria knows who is speaking
Last updated August 17, 2026
Audria separates voices automatically as it listens. There is no enrollment step, and nobody has to read a passage into your phone. Voices Audria has not met yet appear as placeholder people, and once you name a voice, that name applies everywhere, including your past memories.
Name a voice once
To name a voice, pick or create a contact on your device. Audria then relabels that person retroactively across your past memories and reports how many were backfilled, so you can see the history update. One naming, applied everywhere, including conversations captured before you got around to it. Because naming uses your device contacts, the person in Audria is the same person in your address book, and creating a new contact covers anyone who is missing.
Still learning this voice
A voice Audria is still learning shows a dedicated state, and it resolves on its own as Audria hears more. You cannot force a name onto a voice in this state. That restraint protects accuracy: a wrong name attached early would spread to every memory that voice appears in, so Audria waits until it is confident before letting you attach one.
Managing people
You can rename a person, remove their identification, or merge two voices that turned out to be the same person. Each person page shows their display name, how many conversations you have had, when you last saw them, and a playable voice sample so you can confirm who the voice belongs to.
Your voiceprint is tied to your account
Recognising voices means Audria holds a voiceprint. Yours is deleted with your account: delete the account, and once the 30 day grace window completes, voiceprints are removed along with everything else held for you. The deleting your data article in this guide covers that flow step by step.