Velvet

Velvet is iZotope's intelligent de-esser and vocal processor, designed to address the endemic problem of sibilance control without sacrificing vocal presence. The plugin separates sibilant frequencies from tonal content through a machine learning algorithm, allowing targeted processing that conventional single-band de-essers cannot achieve. Upon activation, the Learn function analyzes incoming audio and automatically configures a six-band dynamic EQ to address problematic frequencies, though users retain full manual control for customization.

The tool's architecture comprises three primary channels: Sibilance for hiss reduction, Tonal for dynamic shaping, and De-Click for transient repair. The dynamic EQ operates intelligently, reducing sibilants only when they exceed a user-defined threshold, preserving natural vocal dynamics across quieter passages. The Tonal section functions as an adaptive tone shaper with eight target profiles - including vintage, modern, and bright signatures - that boost or attenuate complementary frequency ranges through independent Lift and Tame controls.

Velvet represents a meaningful advancement over static de-essers and broad spectral processors. Its separation of sibilance from tonality allows engineers to suppress harsh frequencies without the hollow, over-processed character typical of conventional approaches. The De-Click feature effectively removes plosives and mouth noise transparently, reducing cleanup time in the editing phase.

The plugin suits vocal-centric workflows across genres, particularly valuable for close-miked recordings, podcast production, and vocal-heavy mixing. While specialized editors exist for detailed transient work, Velvet's integrated approach makes it efficient for rapid vocal refinement. The fifty included presets provide reliable starting points, though the interface rewards hands-on tweaking from experienced users seeking transparent, musicality-preserving results.