Diffusion-Based High-Fidelity Singing Voice Synthesis With Expressive Performance Control. JAAIR 2026, 5 (1).