A voice profile is a saved sample of someone’s voice. When you diarize a transcript, TranscribeX compares each speaker with your profiles and uses the person’s name and color when it finds a match. Voice profiles are a Pro feature.
Create a voice profile
You make profiles from a transcript you’ve already diarized.
- Open a finished job and run diarization. See speaker diarization.
- In Overview › Speakers, rename each speaker to the person’s real name. Use the play button to check who is who.
- Tick the checkbox of each speaker you want to save, and untick the rest.
- Click Save as Voice Profiles. The app confirms how many profiles it saved.
Only speakers that have lines in the transcript and came from TranscribeX’s own diarization can be saved.
Tip: “You can save the same person multiple times to improve matching accuracy.” Save the same name again from other recordings (different rooms, microphones or calls). Samples with the same name are grouped into one profile, and the number next to the profile shows how many samples it has.
Manage profiles
Open Settings › Speakers. Under Voice Profiles you can:
- Rename a profile by editing its name field. This renames every sample in it.
- Change its color with the color well. The color is used when the profile is matched.
- Play a sample to check the voice.
- Delete a profile with the minus button.
Groups
Groups keep a long list tidy, for example by team or client. Click New Group, name it, then use the group menu on each profile to move it in or out (None means Ungrouped). Removing a group keeps its profiles and moves them back to Ungrouped. Groups are for organizing only: diarization always uses all saved profiles.
Match levels
Each profile has a match level that controls how similar a voice must be before it’s matched automatically. Click Advanced in Settings › Speakers to show the picker on each profile.
| Level | Minimum similarity | Use it when |
|---|---|---|
| Strict | 0.8 | The voice is often confused with other people. Only very strong matches are used. |
| Normal | 0.5 | Most profiles. This is the default. |
| Low | 0.3 | The same speaker is often missed. Expect more false matches, so review results more carefully. |
If several profiles pass their level, the most similar one wins. If none pass, the speaker stays as Speaker N.
Manual speakers
Manual Speakers, in the same pane, are names and colors without a voice sample. Click Add Speaker, then type a name and pick a color. They’re never matched automatically, but they appear in the person menu in a job’s Speakers panel, so you can apply a name and color quickly. Manual speakers are useful for people you don’t record often.
Where profiles are used
- Diarizing a job, by hand or automatically after a meeting.
- Bulk transcription with diarization turned on.
- Live transcription with Diarize Live Transcription on. Live diarization uses your profiles to name speakers as they talk. See live transcription.
- Chat: the Who You Are setting lets you pick your own saved speaker, so the AI knows which lines are yours. See Chat.
Voice profiles FAQ
Why wasn’t a saved speaker recognized?
The voice wasn’t similar enough for the profile’s match level. Save more samples of that person, or set the profile to Low.
Why is the wrong name applied to a speaker?
Two voices are close. Set the wrong profile to Strict so it only matches strong similarities.
Are voice profiles stored online?
No. Profiles are saved in TranscribeX on your Mac.
More in Speakers
All help articles · Still stuck? Email support@transcribex.io.