AI Creates Sheet Music for All Skill Levels

Jingwei Zhao, a former PhD student at the National University of Singapore's Sound and Music Computing Lab, has dedicated years to creating artificial intelligence systems that can automatically arrange music. His research focuses on making music accessible by adapting existing songs for different instruments, ensembles, and skill levels, ensuring the core recognisable elements remain intact.

Zhao's notable contributions include the AccoMontage system for piano accompaniment arrangement and the Beat Transformer model for audio beat tracking. These have been presented at top conferences like ISMIR and NeurIPS, with his work also adopted in creative tools and research.

His unique background as a trained musician and baritone, combined with his machine learning expertise, allows him to evaluate AI-generated music from a performer's perspective. This dual fluency has led to significant roles, including serving as a task captain for MIREX and reviewing for major AI conferences.

Zhao has joined Songscription, a US-based education technology startup, to bring his AI arrangement research to a wider audience. The company's mission is to empower musicians worldwide to play, share, and learn songs. Songscription has developed AI that transcribes audio into sheet music, and Zhao's work will help learners of all levels access music tailored to their abilities.

His most recent paper, accepted to ISMIR 2026, demonstrates that his move to industry has not slowed his research output. Zhao believes AI can provide a bridge between the music people love and the music they can play, democratizing access to music education.