Expressive Performance

Expressive performance in music focuses on computationally modeling and generating musically nuanced performances that capture the subtleties of human interpretation. Current research emphasizes developing generative models, often employing transformer networks or parameter-controlled approaches, to synthesize realistic performances from symbolic representations (e.g., MIDI) or audio recordings, incorporating aspects like articulation, dynamics, and phrasing. This work is crucial for advancing music information retrieval, automated music transcription, and creating interactive musical systems, such as intelligent accompaniment systems, that respond dynamically to human input. Challenges remain in accurately evaluating generated performances and ensuring generalization across diverse musical styles and performers.