Speaker labeling is only seemingly possible in an automated way once the transcription finishes. I would like to be able to invoke it again.