Audio Texture

Audio texture research focuses on generating and manipulating soundscapes characterized by their statistical properties rather than individual notes or events. Current efforts leverage generative adversarial networks (GANs) and StyleGANs, often employing example-based or soft-label conditioning to achieve control over generated textures and enable morphing between different sonic characteristics. A key challenge lies in developing robust evaluation metrics that accurately capture perceptual similarity and sensitivity to variations in texture parameters, with recent work exploring deep-feature based approaches like Gram matrix comparisons and cochlear-model analyses. Improved control and evaluation of audio textures will have significant implications for music composition, sound design, and virtual/augmented reality applications.