Tracks who spoke when, including when voices overlap. Up to eight speakers. 100M parameters. Connections No outgoing connections.
