Phoneme detection and spectral de-essing
smart:deess uses a trained neural network to identify phonemes such as S, Z, Sh, Ch, T and P in real time. Each sibilant is detected from its exact start to its exact end, and the whole sound is processed rather than only its loudest part. Each phoneme type is handled independently, so an "Ess" is treated differently from a "Ch" or a "T".
Beyond a broadband gain component, the plugin applies spectral processing to shape harsh sounds. Sonible notes that dull or lisping sibilants can also be revived with this processing.
