Generate speaker‑labeled transcript from an audio file