Interactive Music Generation With Feedback-Guided Regeneration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing music generation technologies using machine learning models often fail to meet user needs due to limitations in model performance and accuracy of user text descriptions, resulting in suboptimal music output.
Innovation Solution
A music generation method that allows users to adjust automatically generated music through an agent-guided process, enabling regeneration based on user input to improve music quality and alignment with user preferences.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If music is automatically generated using a machine learning model based on text description, then music generation efficiency is improved, but music quality and user satisfaction deteriorate
Solution Approach 1:
The patent implements a feedback mechanism where the generated music is evaluated by an evaluation model, and the evaluation results are used to guide optimization of the generation model. This closed-loop feedback system allows the system to learn from its outputs and progressively improve music quality while maintaining efficient automatic generation capabilities.
Solution Approach 2:
The system performs self-optimization by automatically evaluating its own generated music and using the evaluation results to train and improve its generation model without requiring constant external intervention. The generation model autonomously adjusts its parameters based on evaluation feedback to enhance music quality.
2Device complexity
If the machine learning model performance is limited, then device complexity is reduced, but music generation accuracy deteriorates
Solution Approach 1:
The patent introduces an evaluation model as an intermediary between the generation model and the final music output. This evaluation model assesses the generated music and provides guidance for optimization, enabling the system to achieve higher accuracy without proportionally increasing the complexity of the generation model itself.
Solution Approach 2:
The system dynamically adjusts the generation process based on evaluation feedback. The generation model iteratively refines its outputs by incorporating guidance from the evaluation model, allowing it to achieve higher accuracy adaptively without requiring a static increase in model complexity.
Data Source
AI summary
The present disclosure relates to a music generation method, music generation apparatus and computer-readable storage medium, and falls into the field of computer technology. The music generation method includes: receiving a first prompt information input by a user, the first prompt information comprising descriptive information of music; generating a first musical work according to the first prompt information, the first musical work comprising a first music; and generating a second musical work according to an adjustment operation of the user on the first musical work, the adjustment operation comprising at least one of an adjustment operation on the first prompt information or an adjustment operation on the first music.