Ki is a neural piano sound synthesizer. It can generate an infinite number of strange, unique sounds that cover various keyboard tones. The user can move through the infinity of sounds of the instrument.
Musicians enjoy analogue synthesisers, and more generally analogue audio technology, for the “grain” in the sound. Digital audio technology, for all its perfection and reliability, is often regarded as sounding “cold” and without personality. What would be “grain” in A.I.-based audio synthesis?
Based on Style-GAN 2, Ki was trained on purposely biased spectral representations of keyboard sounds. The result is a model that generates sounds with a strong personality. A creative mind can turn this technical “defect” into a stylistic element.