Traditional granular synthesis algorithms operate in the time-domain, scattering small samples of an audio stream by means of windowing and delay. spectral_granulation.ck extends this concept to the frequency domain. Real-time FFT operations (known also as STFT operations) involve analyzing successive windowed frames of an audio stream. These analysis frames are transformed in to spectral data (FFT "bins"), which are manipulated and resynthesized to the time domain via the inverse FFT. Spectral granulation is achieved by creating a circular buffer of analysis frames, the elements of which are delayed by a stochastically controlled amount before being resynthesized and added to the audio stream. By itself, this operation is no different than regular granulation except for the the contraints imposed by the FFT frame size and granularity (hop size). However, spectral granulation affords the ability to perform addiontional processing on these frames in the frequency domain. spectral_granulation.ck filters each frame to a narrow bandwidth, so that the 'scattering' applies to both spectral and temporal regions of the sound.