I know this will require another model, but it would be nice to have speaker identification to make it easier to transcribe podcasts.