GrainSynth: A Generative Synthesis Tool Based on Spatial Interpretations of Sound Samples

This paper proposes a generative design approach for the creative exploration of dynamic soundscapes that can be used to generate compelling and immersive sound environments. A granular synthesis tool is considered based on the perceptual self-organization of sound samples by utilizing the t-Stochastic Neighboring Embedded algorithm (t-SNE) for the spatial mapping of sonic grains into a 2D space. The proposed system was able to relate the visual stimuli with the sonic responses in the context of the generic gestalt principles of visual perception. According to user evaluation, the application operated intuitively and also revealed the potential for creative expressiveness both from the user’s perspective and as a standalone, generative synthesizer.