Speaker Recognition

Parrot Scribe features advanced speaker recognition that identifies and tracks individual voices across your sessions. This allows you to maintain a consistent record of who said what, even in complex multi-person conversations.

Persistent Recognition

Unlike basic transcription tools that only identify speakers within a single file, Parrot Scribe uses persistent speaker recognition. Once a speaker is identified and labeled, the app will recognize that same voice in future sessions.

This recognition happens entirely on your device. Speaker embeddings—mathematical representations of a voice—are stored securely and never leave your Mac.

Speaker Labels

When a new speaker is detected, they are assigned a generic label (e.g., "Speaker 1"). You can rename these labels at any time to reflect the speaker's actual name.

1. Open a session in the Session Viewer.
2. Click on a speaker label in the transcript.
3. Enter the name and press Enter.

Once updated, this name will be used for that speaker across all past and future sessions where their voice is recognized.

Managing Speakers

Parrot Scribe provides two ways to manage speaker data, depending on your needs for privacy and data retention.

Remove Speaker

"Remove Speaker" is a reversible action. It removes the speaker's name and label from your active speaker list, but retains the underlying voice data. If that person speaks in a future session, they will be recognized again, but will be assigned a new generic label.

Forget Speaker

"Forget Speaker" is a permanent delete action. This action:

• Purges the speaker's voice embeddings from your device.
• Deletes sessions where that speaker appears.

This action is irreversible. If the person speaks again in the future, they will be treated as a completely new, unknown speaker.

Re-enrollment

After you forget a speaker, that person appears as a new speaker on the next encounter.