Research preview of Quorlyn's most expressive text-to-speech model, with control over emotion, emphasis, and pacing.