I don't think they do it automatically, though. I think you need to piece apart the transcript in their tool to decide which voice to use where.