How to de-ess a vocal
Set the de-esser's frequency to where the singer's "s" is loudest, usually 5 to 9 kHz, and lower the threshold until it takes 3 to 6 dB off the esses and does nothing on vowels. If the singer starts to lisp, you've gone too far.
What sibilance is
"S", "sh", "ch" and "t" are bursts of noise made with the tongue and teeth. Their energy is concentrated in a band a few kilohertz wide. Condenser microphones, bright EQ and compression all make that band louder relative to the rest of the voice, and it ends up painful on earbuds.
| Voice | Where the "s" usually peaks | "Sh" sits lower |
|---|---|---|
| Lower voices | 4.5 to 6.5 kHz | 2.5 to 4 kHz |
| Higher voices | 6 to 9 kHz | 3 to 5 kHz |
Setting it up
- Loop a line with several esses in it.
- If your de-esser has a listen or monitor mode, switch it on and sweep the frequency until the "s" is at its harshest. Otherwise sweep a narrow EQ boost to find the spot, then remove the boost.
- Choose split-band mode if it's offered. It turns down only the sibilant band. Wideband mode ducks the whole vocal, which is smoother on some voices and audibly dips on others.
- Lower the threshold until the gain reduction meter shows 3 to 6 dB on esses and stays at zero on vowels.
- Bypass and compare. The words should still be clear. "Yes" shouldn't turn into "yeth".
- Listen on earbuds or a phone speaker. That's where sibilance bothers people.
When a de-esser isn't enough
One or two terrible esses. Fix them by hand. Cut the "s" into its own clip and drop its gain 4 to 8 dB. That's cleaner than tightening the de-esser for the whole song.
Everything is harsh, not just esses. That's an EQ problem. Look for a build-up between 2.5 and 5 kHz and cut 2 to 3 dB with a medium-width bell before you blame sibilance.
The reverb is spitting. Sibilance hitting a bright reverb produces a splash on every "s". Put a de-esser or a low-pass at 6 to 8 kHz on the reverb send, and leave the dry vocal alone.
Don't overdo it
A de-essed vocal measures duller, and that's the point, but past about 8 dB of reduction listeners hear a lisp or a vocal that sounds like the singer has a cold. If you need that much, the vocal was EQ'd too bright upstream. Back off the high shelf by 2 dB and try again.
Quick answers
- What frequency should a de-esser be set to?
- Usually 5 to 7 kHz for lower voices and 6 to 9 kHz for higher voices. Sweep to find where the "s" is loudest for your singer.
- Does the de-esser go before or after the compressor?
- After is the common choice, because compression makes sibilance louder and the de-esser then catches the result. Before works when the compressor is reacting to the esses.
Audizer does this when you ask.A Mac DAW with an engineer built in. Say what you want and it makes the edit. Free download.
Get Audizer