When de-essing, smart:deess analyzes the signal’s actual content – not just the level of one or multiple frequency bands – to make decisions. Since each person has their own voice characteristics, smart:deess asks you to playback some audio to learn a short vocal fingerprint. This vocal fingerprint – we also call it voiceprint – is like a fingerprint of the vocal characteristics of the analyzed person. Once a voice has been analyzed, smart:deess adapts its internal processing to match the observed characteristics.
When working with a new speaker or singer, you should therefore always re-analyze the signal to make sure that the processing adapts accordingly.