Name the speakers
Listen to a sample clip of each detected speaker, then enter a display name. Saving will regenerate the summary using the new names.
If diarization split the same person into two speakers, use "Merge into" on one of them to fold it into the other.