An end-to-end framework for multi-speaker transcription that jointly models who spoke, when, and what.
