Grain Selection Method for Granular Synthesizer Using Image Input

Abstract

We have developed a system to automate the process of preparing sound sources for granular synthesizers, and to associate the impressions of the user's input images with the frequency bands of the preset sound sources, in order to improve convenience and provide new inspiration.The system consists of a Python script that selects grain presets based on similarity using RGB histograms and sends the presets using OSC, and a synthesizer patch using PureData that can read the preset sound sources and change the pitch of the pronounced tone using MIDI. In order to link the user's impression of the input image with the frequency characteristics of the grain preset, we created a questionnaire to evaluate the impression of the image and the frequency band of the preset. We proposed a method to select the grain with the frequency band whose impression is close to the input image.