Methods of Generative Audio Synthesis
Sciences and Mathematics, College of
Chemistry and Physics, Department of
BURS Faculty Advisor
Abstract. With the rise of AI technologies, questions of “how can we apply this where it hasn’t been already” becomes imperative. There have been many developments in the use of neural networks for generating audio, but this exploratory discussion centers around the robustness of using a neural network to approximate a manifold upon which data exists. The existence of a manifold in the data implies that there is a mapping back into the input space, thus making the ability to go from a relational space back to the audio space very straightforward. In this way, we believe it may be a key to constructing neural network-based generative audio synthesis.
Ortner, Max and Hawley, Scott, "Methods of Generative Audio Synthesis" (2023). Belmont University Research Symposium (BURS). 332.