Voice memory

Rename a speaker once and Callmora remembers their voice, so the same person is named for you in future calls — no re‑tagging, and nothing about their voice leaves your Mac.

What it does

Callmora already separates who said what and labels each voice as Speaker 1, Speaker 2, and so on. Voice memory adds the next step: when you give one of those speakers a real name, Callmora learns what that person sounds like and matches them the next time they're on a call.

What it stores is a voice signature — a short list of 256 numbers derived from the audio, not a clip of anyone speaking. Signatures live in a small database on your Mac, kept apart from your notes, and you stay in control of them.

Teach Callmora a voice

There's nothing extra to set up. Teaching a voice is the same rename you'd do anyway:

Open a call that has finished processing and find a speaker labelled Speaker 1, Speaker 2, and so on.
Rename them to a real name — for example, change Speaker 2 to Alex. The name applies across the whole call right away.
That's it. From that call on, Callmora has a signature for Alex and looks for his voice in new calls.

Any call works — a live recording or one you imported. You only teach a voice once; renaming the same person again just keeps the signature current.

Automatic names and suggestions

On a new call, Callmora compares each speaker to the voices it already knows:

A confident match is named for you. When a speaker clearly matches Alex, they arrive labelled Alex — in the transcript, the notes, and your to‑dos.
A close-but-unsure match is offered as a suggestion. When Callmora thinks it recognizes someone but isn't sure, the speaker editor shows a "Sounds like Alex — apply" prompt. Accept it to use the name, or ignore it and rename by hand.

Voice memory never overwrites a name you've set yourself, and it never applies to Me — your own voice stays yours. If a suggestion or an automatic name is wrong, rename the speaker as usual, then clean up the remembered voice in Settings → Voices (below).

Managing remembered voices

Everything Callmora remembers lives in Settings → Voices. There you can:

See every remembered voice by name.
Delete one you no longer want matched.
Forget all of them at once.

If the wrong person gets learned — say two colleagues sound alike and Callmora mixes them up — delete that voice in Settings → Voices and rename the speaker again on a clear call, so a fresh signature replaces the old mistake. "Erase all data" removes remembered voices along with everything else.

Privacy

A voice signature is a numeric vector — 256 numbers — not a recording, and it's derived on your Mac. Signatures sit in their own local database and are never included in the iCloud sync of your notes.

If you want them on your other Macs, turn on Settings → Voices → "Sync voices via iCloud" (off by default); that mirrors the signatures to your own private iCloud database, and switching it back off deletes the cloud copies. You can remove individual voices or all of them at any time, and "Erase all data" clears them too.

For the full picture, see the privacy policy.

Good to know

Learning starts with new calls. Voice memory kicks in for calls processed after this update. Older calls still rename the same way — that just isn't learned from.
Local or cloud, either works. Signatures are built on your Mac whether you transcribe on‑device or in the cloud.
Imported files count too. Rename a speaker on a call you imported and Callmora learns the voice just as it would from a live recording.

← Back to all docs