Computational Analysis of Jazz Improvisation: Integrating CREPE Pitch Detection with Language Model Feedback
DOI:
https://doi.org/10.58445/rars.4094Keywords:
Jazz improvisation, artificial intelligence, CREPEAbstract
Jazz improvisation is inherently subjective, which makes it difficult for musicians
to receive consistent, actionable, and quantitative assessments of their performances.
This paper presents a method of analyzing monophonic jazz solos computationally
through the combination of pitch estimation and large language model (LLM) reasoning.
After transcribing the uploaded solo by using the Convolutional Representation for Pitch
Estimation (CREPE) model, the note sequence is placed alongside the provided chord
timeline. This alignment allows the LLM to generate structured feedback on phrase
structure, harmonic alignment, and motif development, as well as provide actionable
advice to improve. A prototype web application using this pipeline was evaluated by
eight musicians of varying experience and skill. While experienced jazz soloists rated
the feedback as highly actionable, beginning musicians often found it difficult to
decipher the musical jargon, thus indicating the necessity of simple and
easy-to-understand language. Currently, this framework does not currently assess
dynamics, articulation, or other expressive characteristics. However, it demonstrates the
potential of integrating pitch estimation and LLMs to generate structured feedback on
jazz improvisations, providing a computational system for improvisational analysis that
has future applications in AI-assisted music education.
References
Biles, John A. “GenJam: A Genetic Algorithm for Generating Jazz Solos.” International
Computer Music Conference (ICMC), July 1994, pp. 131–137.
Hein, Ethan. ““Humans Will Be Doing All the Serious Music Transcription for the
Foreseeable Future”: Songscription Review.” MusicRadar, 8 Dec. 2025,
musicradar.com/music-tech/humans-will-be-doing-all-the-serious-music-transcript
ion-for-the-foreseeable-future-songscription-review. Accessed 1 May 2026.
Johnson-Laird, Philip N. “How Jazz Musicians Improvise.” Music Perception, vol. 19, no.
, 2002, pp. 415–442, doi.org/10.1525/mp.2002.19.3.415.
Jordanous, Anna. “Emovectors: Assessing Emotional Content in Jazz Improvisations for
Creativity Evaluation.” 2025 IEEE International Conference on Big Data, 2025,
pp. 5003–5006, doi.org/10.1109/BigData66926.2025.11402279.
Keller, Robert M., and David R. Morrison. “A Grammatical Approach to Automatic
Improvisation.” Proceedings of the Sound and Music Computing Conference,
July 2007, pp. 330–337.
Kim, Jong Wook, et al. “CREPE: A Convolutional Representation for Pitch Estimation.”
ArXiv, 2018, arxiv.org/abs/1802.06182.
Pachet, François. “The Continuator: Musical Interaction with Style.” Journal of New
Music Research, vol. 32, no. 3, 1 Sept. 2003, pp. 333–341,
doi.org/10.1076/jnmr.32.3.333.16861.
Riley, Xavier, and Simon Dixon. “Reconstructing the Charlie Parker Omnibook Using an
Audio-To-Score Automatic Transcription Pipeline.” ArXiv, 2024,
arxiv.org/html/2405.16687v1.
Thom, Belinda. “Interactive Improvisational Music Companionship: A User-Modeling
Approach.” User Modeling and User-Adapted Interaction, vol. 13, no. 1, Feb.
, pp. 133–177, doi.org/10.1023/a:1024014923940.
Wiggins, Geraint A. “Searching for Computational Creativity.” New Generation
Computing, vol. 24, no. 3, Sept. 2006, pp. 209–222, doi.org/10.1007/bf03037332.
Yu, Guo, et al. “Recent Advances in Artificial Intelligence For Music Education.”
Transactions on Artificial Intelligence, vol. 2, no. 1, 27 Feb. 2026, pp. 39–53,
doi.org/10.53941/tai.2026.100004.
Downloads
Posted
Categories
License
Copyright (c) 2026 Research Archive of Rising Scholars
This work is licensed under a Creative Commons Attribution 4.0 International License.
