Vocal sibilance vs harshness: diagnose before de-essing

Loop a phrase with a strong consonant and a held vowel. Mark exactly what makes you want to turn the vocal down, then use this first-pass decision:

Listen to the recording with later effects bypassed, then restore the chain one stage at a time at comparable levels. Compression can change the relative prominence of quieter syllables; bright EQ and harmonic processing can make an existing edge more noticeable. A distorted recording and an overdriven insert require different decisions from a clean but sibilant performance.

For a hypothetical vocal, suppose the raw esse is acceptable but becomes detached after a high shelf. Reduce that shelf first. If the phrase then feels too dull overall, use a smaller lift with selective consonant control. If the edge appears only after saturation, compare less drive before adding another corrective plugin after it.

Use detector audition briefly to find the consonant energy. Set a limited reduction range, then adjust detection until difficult consonants trigger it and comfortable words mostly pass. Return to full-band listening to judge pronunciation.

Pro-DS documents two separate choices: Single Vocal versus Allround detection, and wide-band versus split-band processing. The latter determines whether detection lowers the whole signal or its high-frequency band. Do not assume every de-esser distinguishes vowels from consonants in the same way. Compare the complete word in each mode; a lower reduction reading is not a better result by itself.

If the vowel itself is hard, try a small static EQ move when the tone is consistently excessive, or a dynamic band when only forceful notes need restraint. Locate the area in context and keep the bandwidth broad enough to solve the audible problem. Avoid turning a boosted solo sweep into a catalogue of every resonance the voice contains.

For one isolated shout, clip gain or automation may be more predictable than a processor acting across the whole song. Listen across the edit for continuity. If reducing one region makes the singer hollow, you may be targeting a useful part of their character rather than the source of fatigue. Back off and revisit level, drive, and masking.

Briefly solo the reverb and delay returns. If their consonant tails are the problem, compare a filtered return, a lower send on the offending syllable, or a feed taken after de-essing. Confirm the send tap point before expecting an insert to change that feed.

Check for a lisp, a dip that continues into the next vowel, or an effect tail that changes color unnaturally. Reduce the treatment if those appear. If an overloaded or unsuitable take is available to rerecord, compare that option; otherwise preserve believable diction rather than removing every trace of edge.