Skip to main content

Merge & Split Speakers

Fix speaker attribution by merging duplicates and splitting combined utterances.

Understanding Speaker Errors​

Over-segmentation: One person split into multiple IDs. Under-segmentation: Multiple speakers combined in one utterance.

Merging Speakers​

Merge when two IDs represent the same person. Click the speaker badge, select "Rename this speaker", and type the target speaker name. All utterances update automatically.

Splitting Utterances​

Split when one utterance contains multiple speakers. Position cursor at the split point, click Split button (scissors icon) or press Ctrl+Shift+Down (Cmd+Shift+Down), then assign the correct speaker to the second part.

Merging Utterances​

Merge consecutive utterances by the same speaker. Select 2 or more utterances, then click the Merge button that appears.

Tips​

Work systematically:

  • Merge duplicate speakers first
  • Then split any combined utterances
  • Finally rename speakers to actual names

Use audio to verify:

  • Listen to confirm speaker identity
  • Click words to jump to audio position
  • Don't guess - verify with your ears

Undo available:

  • Press Ctrl+Z (Cmd+Z) to undo changes
  • Use immediately if you make a mistake

Next Steps​