METEOR: Melody-aware Texture-controllable Symbolic Orchestral Music Generation

2024-09-18

arXiv.org

AI Summary

This study addresses the dual challenges of low melodic fidelity and uncontrollable accompaniment texture in multi-track symbolic orchestral music generation for Western homotonic textures. We propose the first generative framework integrating high-fidelity melodic constraints with fine-grained, controllable textural attributes—including density, rhythmic patterns, and pitch register. Our method employs a Transformer-based multi-condition joint modeling architecture comprising a dedicated melody encoder, a text-driven texture embedding module, and a hierarchical attention mechanism, augmented by contrastive learning to ensure melodic consistency. The framework supports bar-level and voice-level textual conditioning for stylistic transfer and accommodates twelve professional texture descriptors. Experiments demonstrate a 23.6% improvement in melodic fidelity over state-of-the-art methods and a 91.4% accuracy in texture control, substantially overcoming the long-standing melodic distortion bottleneck in symbolic orchestral music generation.

Technology Category

• Natural Language Processing: Generation
• Machine Learning: Deep Generative Models & Autoencoders
• Search and Optimization: Mixed Discrete/Continuous Search

Application Category

• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Western music is often characterized by a homophonic texture, in which the musical content can be organized into a melody and an accompaniment. In orchestral music, in particular, the composer can select specific characteristics for each instrument's part within the accompaniment, while also needing to adapt the melody to suit the capabilities of the instruments performing it. In this work, we propose METEOR, a model for Melody-aware Texture-controllable Orchestral music generation. This model performs symbolic multi-track music style transfer with a focus on melodic fidelity. We allow bar- and track-level controllability of the accompaniment with various textural attributes while keeping a homophonic texture. We show that the model can achieve controllability performances similar to strong baselines while greatly improve melodic fidelity.

Problem

• Generates orchestral music with melody-aware control
• Enables texture-controllable accompaniment in homophonic style
• Improves melodic fidelity in symbolic music transfer

Innovation

• Melody-aware symbolic orchestral music generation
• Bar- and track-level controllable accompaniment
• Homophonic texture with melodic fidelity