Generative Model for Musical Arrangement Generation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional methods for generating musical arrangements are costly and limited in generating varied data, as they rely on prescribed algorithms that may not match the original performance information, resulting in deviations from the original piece and inability to produce non-uniform arrangement data.

Innovation Solution

A method using a computer to acquire target musical piece data with performance information and meta information, employing a generative model trained by machine learning to generate arrangement data that aligns with the meta information, allowing for the automation of arrangement generation and production of varied data.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Extent of automation

If a prescribed algorithm is used to generate arrangements from performance information, then the generation process can be automated, but the arrangement may deviate from the original piece and appropriate data may not be generated

Engineering Contradiction:

Improveautomation of arrangement generationVSAvoidaccuracy of arrangement matching original piece

Core Design Contradiction:

Extent of automationVSReliability

Solution Approach 1:

The patent transforms the arrangement generation process from using fixed prescribed algorithms to using a machine learning model with adjustable parameters that can be trained on specific musical styles and characteristics. This allows the system to adapt its behavior based on the input performance information while maintaining automation, resolving the contradiction between automated generation and accuracy.

Inventive Principle:

Principle #35Parameter changes

2Extent of automation

If a prescribed algorithm is used to generate arrangements, then the process can be automated, but only uniform arrangement data can be generated making it difficult to produce varied arrangements

Engineering Contradiction:

Improveautomation of arrangement generationVSAvoidvariety of arrangement data

Core Design Contradiction:

Extent of automationVSAdaptability or versatility

Solution Approach 1:

The patent introduces dynamic adaptability by training the machine learning model on diverse training data representing different musical styles and arrangement types. The model can dynamically adjust its output based on the characteristics of the input performance information and desired arrangement style, enabling automated generation of varied arrangements rather than uniform outputs.

Inventive Principle:

Principle #15Dynamics

3Reliability

If manual operations are used to generate musical scores, then appropriate and varied arrangement data can be generated, but the cost of generating the musical score increases

Engineering Contradiction:

Improvequality of arrangement dataVSAvoidcost of generating musical score

Core Design Contradiction:

ReliabilityVSProductivity

Solution Approach 1:

The patent uses machine learning to learn from existing musical arrangements (training data) and generate new arrangements by applying learned patterns rather than manual copying. The system replicates the quality of expert arrangements by training on diverse musical data, then automatically generates varied arrangements without requiring manual operations for each new piece, reducing cost while maintaining quality.

Inventive Principle:

Principle #26Copying

Data Source

PatentUS20220383843A1Arrangement generation method, arrangement generation device, and generation program

Publication Date: 2022.12.01 YAMAHA CORP

AI summary

An arrangement generation method executed by a computer includes acquiring target musical piece data that include performance information that indicates a melody and a chord of at least a part of a musical piece and include meta information that indicates characteristics of at least the part of the musical piece, generating, from the acquired target musical piece data, by using a generative model trained by machine learning, arrangement data obtained by arranging the performance information in accordance with the meta information, and outputting the generated arrangement data.