In Fall of 2015 I developed a project which explores the visual representation of one of my favorite audio synthesis techniques, granular synthesis. A granular synthesizer generates new audio from a sample piece of audio by slicing the sample into slices, or “grains”, of very short length. Typically, a randomized selection of grains from different parts of the sample will be replayed in an overlapping fashion, generating a new sound and timbre distinct from that of the original sample.
My visualization, developed in the Pd audio programming language, prompts the user to select an audio sample and, additionally, an image to correspond with the audio for the subsequent granular process. As new sounds are generated from the granular synthesis, a visual representation of the four most recent grains are superimposed and displayed on the screen.
The visualization concept is as follows: each grain image is scaled to fill the application window exactly. We suppose a long “grain” the length of the original sample should be represented by the original undistorted image. A grain half the length of the original sound sample, and starting at its beginning, should be represented by a vertically centered square section of the original image, with length 1/2 of the original image. In this way, smaller audio grains are represented by equally “smaller” visual grains, enlarged to fill the same visual space. Additionally, the process of pitch shifting in the audio domain is represented by horizontal stretching and compression in the visual domain.