As part of our ongoing effort to improve transcription accuracy, we've updated our speaker detection models.
While this has significantly improved the overall quality of your transcripts, you might notice that the new model is more sensitive and sometimes identifies short conversational fillers (like "yeah," "okay," or "hmm") as a separate speaker.
This can occasionally result in your transcript showing one more speaker than was actually present.
Don't worry! This won't impact the quality of your insights, and you have two simple options to manage it.
Option 1: Ignore the Extra Speaker (Quickest Method)
If the discrepancy is minor and doesn't affect your ability to read the transcript, you can simply ignore the extra speaker label (e.g., "Speaker F"). You can still name the correct speakers and proceed to create content and insights as usual.
Option 2: Regenerate the Conversation with an Accurate Speaker Count
If the speaker count is significantly off or you prefer a perfectly clean transcript, you can use our Fix Speaker Count tool to re-process the audio. This gives you more control and helps our AI deliver a more accurate result.
Here’s how to do it:
From within your conversation transcript, click the Label [X] Speakers button in the top action bar.
The Name Your Speakers panel will open on the right. At the bottom of this panel, click the Fix Speaker Count button.
A "Fix Speaker Count" window will pop up. Here, you can guide our AI by setting a range for the number of speakers you expect.
Minimum Speakers: Enter the lowest possible number of speakers.
Maximum Speakers: Enter the highest possible number of speakers.
Pro Tip: For the best results, be as specific as possible.
If you know exactly 6 people spoke, set both Minimum and Maximum to 6.
If you are unsure but think it was between 5 and 7 people, set the range accordingly.
Click the Regenerate button. The system will re-process your audio file with the new speaker parameters. You will see a progress indicator as it completes the transcription and summarisation.
What to Expect After Regenerating
A New Version is Created: A new, updated version of your conversation will be generated and will appear in your Conversation Library. The title will often have a (2) appended to it, which you can easily rename.
Review and Label: Once processing is complete, you can review the new transcript and label your speakers. The speaker count should now be accurate based on the range you provided.
We hope this feature gives you even more control over the accuracy and presentation of your important conversations