Speaker Recognition

Version: This guide describes Parrot Scribe 1.1.0.

Parrot Scribe tells speakers apart in each recording and lets you name them. People you have not named appear with generic labels such as "Speaker 1". Recognizing the same person in later recordings is a separate, optional feature that is off until you turn it on.

Before You Begin

To review speakers after a session, keep its audio. In Settings > Output, leave Retain audio recordings on and choose how long to keep audio under Keep audio for. Parrot Scribe needs the retained recording to play each speaker's sample and to run speaker analysis again.

Speaker analysis runs when a recording finishes. Named speakers do not appear in the live transcript during recording.

Name Speakers in a Session

Open the session in Library and choose Identify Speakers.

1. Review each detected voice and, if needed, play its sample.
2. If two voices belong to one person, select both and choose Same Person. Undo Grouping reverses the last grouping.
3. Enter a name, or choose a person you named before.
4. Select Save Changes.

Names saved this way apply to this session. They do not save a voice profile, and they work whether or not voice recognition is on.

To mark your own voice, choose This is me on that speaker. If your voice is already marked elsewhere in the session, Parrot Scribe asks before merging the two.

Voice Recognition

Voice recognition creates a voiceprint from each speaker in a recording and compares it with saved voice profiles, so Parrot Scribe can suggest names in later recordings. Voiceprints stay encrypted on your Mac.

Voice recognition is off by default, including after an upgrade from 1.0.x. To turn it on:

1. Open Settings > Transcription.
2. Turn on Recognize Saved Voices.
3. Read the notice and choose Enable Voice Recognition.

Make sure the people you record know about this use and that you have any permission required where you work. Your confirmation in the app is not a substitute for their consent.

While recognition is on, Identify Speakers can offer Apply [n] Matches for speakers that match saved voices. Review the suggestions before applying them.

Turning Recognize Saved Voices off stops new voiceprints and matching. It does not delete saved voice profiles. To remove a profile, see Managing Speakers.

Save a Voice Profile

A voice profile lets Parrot Scribe recognize a person in later recordings. Saving one is always a separate choice:

1. Turn on voice recognition, as described above.
2. In Identify Speakers, name the speaker and check Remember [name]'s voice. The box starts unchecked.
3. Select Save Changes, then review the Save Voice Profiles? confirmation and choose Save Voice Profiles.

Cancel closes the confirmation without saving your changes. To save session-only names, uncheck Remember [name]'s voice and select Save Changes again. A speaker with too little speech shows Too little speech to remember this voice instead of the checkbox. Remembering a voice for someone you already saved adds this recording's sample to their profile.

Saved names are not applied to past sessions automatically. Voice profiles imported from 1.0.x need new samples before they can match.

Managing Speakers

Open a person in the Speakers section of Library.

• Rename: click the name to change it everywhere it appears.
• Merge into [name]: combines two entries for the same person, including their session links and any saved voice profiles. Parrot Scribe shows both names and asks you to confirm.
• Remove Speaker: removes the person from your list and deletes their voice profile. Transcript text stays in its sessions.
• Delete Speaker and Sessions…: deletes the person, their saved voice and every session attributed to them from this library, including other speakers' text and retained audio. This cannot be undone here. Recovery copies or backups outside the library may still contain the data.
• Export Speaker Summary…: saves a JSON file with the person's saved details and known session links. It excludes transcript text, audio, raw voiceprints and other speaker profiles. It does not find every mention of a person; review session titles before sharing it.

Run Speaker Analysis Again

In Identify Speakers, choose Re-run Speaker Analysis… to analyse the session's recording again. Names given in that session are cleared. Saved voices are matched again only if voice recognition is on. Your transcript and recording stay unchanged.