Ghira
Docs menu

Using Ghira

Speakers and voices

Naming and fixing speakers, People, your own voice profile

Ghira tells speakers apart while you record and keeps the labels in the transcript and the notes. You can name them live or later.

How speakers are labeled

Name a speaker

You can rename during the recording or in a saved meeting.

  1. Select the speaker. Ghira asks Who is this?
  2. Type a name, or pick someone already in People.
  3. The name applies to every line of that speaker, and to the notes.

Play 3 s sample plays a short clip of the voice, so you can check who it is. Leave the name empty to go back to "Speaker N".

Names you give feed the custom vocabulary, so the final transcript spells them correctly. See Learned from names you gave speakers in Settings → Languages.

Fix a mistake

Open Speaker options for a speaker to get the speaker panel. It offers these actions:

In a Room recording, This is me marks a speaker as you, and Not me undoes it.

People

The People screen lists everyone you have named, how many meetings you shared, and open action items. From there you can:

Your voice profile

A voice profile lets Ghira recognize a voice without asking. It is a numeric summary of how someone sounds. It cannot be played back as audio, it is stored encrypted on your Mac, and it is never uploaded. See Privacy.

You can save a profile for yourself:

  1. Download the voice model in Settings → Models.
  2. Open Settings → Privacy and find Your voice.
  3. Choose Record your voice again… and read the passage aloud. It takes about 20 seconds. A new recording replaces the old profile.

With a profile, Ghira can suggest and apply Me in Room recordings. Choose Delete my voice data… at any time; your name stays in the notes.

Other people's voices

Saving other people's voice profiles is switched off in this version. You can still name anyone, and the name is kept in your meetings and in People. Ghira does not save their voices, and Settings → Privacy says so next to Learn voices from confirmed speakers.

Ghira also never learns a voice from a sensitive meeting. See Recording.

Limits

Edit this page on GitHubLast updated