How transparent is your de-esser? Here's how to test it

Here's a question almost no de-esser answers: when it removes the harsh esses, how much of the rest of the vocal does it damage? Every plugin claims to be "transparent." Almost none publishes a number. And "transparent" without a measurement is just a nice word. This is how to actually test a de-esser's transparency yourself, what a good result looks like, and why we publish ours.

Key Takeaways

• "Transparent" is a measurable claim, not an adjective — you can test it.
• The test: measure how much a de-esser changes non-sibilant sound (vowels, air) while it's working.
• A transparent de-esser leaves the non-sibilant band nearly untouched — a fraction of a decibel.
• We publish Sibilance's number (±0.04 dB); most de-essers publish nothing, so you're trusting an adjective.

What "transparent" actually means

A de-esser has one legitimate job: reduce sibilance. Every decibel it takes off anything else — the vowel's air, the body of the voice, the presence — is collateral damage. Transparency is a measure of that collateral: how little the de-esser disturbs the parts of the vocal that aren't sibilant, while it's busy reducing the parts that are.

A transparent de-esser pulls the ess down and leaves everything else where it was. A non-transparent one dulls the whole top end to get there. Both remove the ess; only one keeps the vocal open. The difference is a number, and you can measure it.

How to test it yourself

You don't need a lab. You need a way to compare the output to the input on material that shouldn't be affected:

1. Feed the de-esser non-sibilant sound. A sustained vowel, a tone, or pink noise in the vowel and air region (say a band around 8 kHz) — sound with no esses in it.
2. Match the levels so you're comparing processing, not loudness.
3. Measure the difference between the de-esser's output and its input in that non-sibilant band, with the de-esser set to a normal amount of reduction. A null test (invert one and sum) or a simple level comparison both work.
4. Read the number. How many decibels did the non-sibilant band move? That's your transparency figure.

A genuinely transparent de-esser moves the non-sibilant band by a fraction of a decibel. If it drops the 8 kHz air by a decibel or two just from being switched on and working, that's the dulling you hear as a lifeless vocal. (Why that dulling happens is in how to de-ess without dulling the vocal.)

What a good number looks like

There's no universal pass mark, but the principle is simple: the smaller the change to non-sibilant sound, the more transparent the de-esser. Tenths of a decibel is excellent. A decibel or more of collateral on the vowel band is where you start hearing the vocal go dull.

We hold Sibilance to a two-sided gate on exactly this: on our test material the vowel-HF band stays within ±0.04 dB of untouched while it removes about 5 dB of sibilance, and the full method and numbers are on the validation page — because a transparency claim should be a measurement you can check, not an adjective you have to trust.

Why almost nobody publishes this

Because measuring it honestly is work, and publishing it invites scrutiny. It's easier to write "transparent" in the marketing copy and move on. When you read a de-esser's page and it asserts transparency with no number behind it, that's the tell: the claim hasn't been measured, or hasn't been measured in a way anyone's willing to show. You're free to run the test above on any de-esser you own and see for yourself.

The short version

Transparency is testable: feed a de-esser non-sibilant sound, match levels, and measure how much it moves the vowel and air bands while it works. Tenths of a decibel is transparent; a decibel or more is dulling. Most de-essers ask you to take "transparent" on faith — we publish the number, and you can run the test on Sibilance's free tier yourself.

Sibilance is the single-authority de-esser this comes from — a real free tier, and a 7-day Pro trial with no card.

Related reading