Real-Time Performance Controllers for Synthesized Singing

A wide variety of singing synthesis models and methods exist, but there are remarkably few real-time controllers for these models. This paper describes a variety of devices developed over the last few years for controlling singing synthesis models implemented in the Synthesis Toolkit in C++ (STK), Max/MSP, and ChucK. All of the controllers share some common features, such as air-pressure sensing for breathing and/or loudness control, means to control pitch, and methods for selecting and blending phonemes, diphones, and words. However, the form factors, sensors, mappings, and algorithms vary greatly between the different controllers.