Qiaoxi Zhang will be presenting the following paper at the TENOR 2025 conference (October 25-27, 2025):
- Zhang, Q., Barthet, M., Xambó Sedó, A. (accepted) “From Shape to Music: Contour-Conditioned Symbolic Music Generation”. Proceedings of the International Conference on Technologies for Music Notation and Representation (TENOR 2025). Central Conservatory of Music Beijing, Beijing, China.
Abstract: We present a novel approach to symbolic music generation that enables users to guide melodic trajectories through abstract contour inputs. This work demonstrates how low-effort, sketch-based controls can influence the expressive shape of generated music, paving the way for more intuitive composition tools. Our framework consists of two main stages. In the first stage, we extract melodic contour from symbolic music using the Ramer–Douglas–Peucker (RDP) algorithm, which simplifies the melodic trajectory while preserving its essential directional shape. Second, a Transformer-based generative model is conditioned on user-specified contour tokens to produce polyphonic continuations that follow the desired melodic shape. To strengthen contour adherence, we introduce a contour-aware loss based on cosine similarity between the target contour vector and the generated melodic motion. Objective evaluations on the MAESTRO dataset indicate that the proposed model closely aligns with target contours while preserving overall musical quality comparable to a baseline Music Transformer. These results highlight the feasibility of high-level contour-conditioned music generation and point toward future applications that integrate contour-based inputs with modalities such as hand gestures in VR, enabling intuitive, real-time human–AI co-composition.
The International Conference on Technologies for Music Notation and Representation is dedicated to issues in theoretical and applied research and development in Music Notation and Representation, with a strong focus on computer tools and applications, as well as a tight connection to music creation. “Technology” in the conference name refers to any mean that may contribute to the notation, representation and/or visualisation of the music and sound, for purposes that may include (but not limited to) music composition, performance, representation, transcription, analysis and pedagogy.