Automatic Tune Designer System Using Deep Learning
Main Article Content
Abstract
The development of methods for automated and audio generation has been a critical area of research in audio signal processing. Existing techniques often struggle to fully capture the main theme of music, highlighting the need for more sophisticated approaches. In response to these limitations, this study introduces a Conditional Variational Autoencoder (CVAE) model, utilizing the advanced capabilities of the Python libraries This CVAE model, developed within a TensorFlow and Keras framework, incorporates convolutional layers and ResNet1D blocks to effectively encode and decode audio signals. It aims to understand and recreate the complex patterns found in music accurately. By proposing this model, we address previous shortcomings in automated music generation, offering a novel solution that significantly enhances the quality and coherence of generated audio. The model's effectiveness is quantified using the Evidence Lower Bound (ELBO) metric, ensuring high-quality audio output and marking a substantial progress in the quest for automated music generation.
Article Details
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.