Annodex logo AnnodexNotes for people who make media
Podcasting

Taming Sibilance on a Podcast Voice Without a Lisp

I recorded a guest last spring whose voice sat politely around -18 dBFS and whose letter S hit -6. On speakers it passed.

A dynamic microphone angled slightly off axis above a seated speaker with headphones, beside a laptop showing a narrow dynamic EQ band near seven kilohertz

I recorded a guest last spring whose voice sat politely around -18 dBFS and whose letter S hit -6. On speakers it passed. On headphones, every plural in the episode arrived as a little blast of steam somewhere around 7 kHz, and after forty minutes of listening my ears were tired in a way that had nothing to do with volume.

The advice you will find in ten seconds of searching is to add a de-esser. That is the last step, not the first, and reaching for it early is how voices end up sounding like they have a lisp. Sibilance is mostly a geometry problem, partly a microphone problem, and only at the end a plugin problem. The order you work in decides whether the fix is invisible.

What you are actually hearing

Sibilance is concentrated high frequency energy from the consonants S, Z, SH and CH, produced by air forced through a narrow gap between tongue and teeth. The energy lands in a band, and where in that band depends on the speaker. Most adult voices I record peak between 5.5 and 9 kHz. Lower voices tend to sit around 5.5 to 7, brighter voices nearer 8 to 9. Children and some female speakers can push past 10.

That number matters because the band is narrow. If you sweep a tight bell boost through 4 to 11 kHz while the person talks and listen for the spot where the hiss turns unpleasant, you will usually find it inside half an octave. Write it down. You now know more about that voice than any preset does.

Microphone choice stacks on top of it. Large diaphragm condensers with a presence lift, which covers most of the affordable studio mics sold to podcasters, add 3 to 5 dB somewhere between 5 and 8 kHz by design, because that lift is what makes a voice sound detailed and expensive on a shop demo. It also doubles the problem you are trying to solve. A broadcast dynamic such as an SM7B or a RE20 is flatter up there, which is why radio studios bought them by the crate and why I have never had to de-ess anybody heavily on one.

Fix the geometry before you touch a plugin

Sibilance leaves the mouth as a fairly directional jet of air. Put the capsule outside that jet and a third to a half of the problem disappears before the signal exists.

What works in my room: mic at roughly the height of the bridge of the nose, tilted down so it aims at the mouth, and the speaker talking across it rather than straight into it. Fifteen to twenty degrees off axis is enough. Any more and you lose body. I keep the lips about 15 cm from the grille on a dynamic, further on a condenser, and I move the mic rather than asking the guest to hold a position they will forget inside two minutes.

A foam windscreen does very little here. A pop filter does less. Both are built for plosives, which are low frequency pressure events, and the mesh is acoustically transparent at 7 kHz where your actual problem lives. The pencil taped across the grille, which people still recommend, helped a P and did nothing measurable for an S in the one evening I bothered to test it.

Two other things move the needle. Dry mouths hiss more, so water on the table and a break at the twenty minute mark genuinely reduces it. And a hard desk directly under the mic sends a reflection back into the capsule with the high frequencies intact. A folded towel under the mic stand is not a studio treatment, but it is free, and it changed more on one kitchen table recording than any processing I added afterwards. The same logic applies to a room that rings, except there you are chasing the low end.

Why the de-esser is not the first tool

Here is where I part company with most tutorials. They put a de-esser at the top of the chain as a default insert, sitting there on every voice at a fixed threshold. That approach is how you get the lisp.

A de-esser is a compressor that only listens to one band. Set wide, it ducks the whole upper range every time an S arrives, which pulls down the air and clarity of the vowel underneath it. Set to 8 dB of reduction, it turns an S into a T. I rarely allow more than 4 or 5 dB of reduction on anything, and on most voices 2 to 3 is plenty.

The second thing tutorials get backwards is the order. Compression raises quiet material, and sibilance is often the quiet material, so a compressor can manufacture a problem that the raw take did not have. If the hiss is in the source, de-ess before the compressor. If the raw file is fine and the processed file hisses, the compressor did it, and the right move is a gentler ratio and a slower attack rather than a second de-esser bolted on afterwards.

Dynamic EQ is the better tool in most cases, and nobody ever recommends it for this. One band, narrow, parked on the frequency you found in the sweep, set to reduce only when the level in that band crosses a threshold. It touches less of the signal than a de-esser does, and the result holds up when the same episode gets normalised to the loudness target your host applies, which de-essing done too hard will not.

A chain that holds up on a real episode

  1. Sweep a narrow boost of about 8 dB through 4 to 11 kHz during speech and note the worst frequency. Remove the boost.
  2. Clip gain the ten or fifteen worst individual S sounds by 3 to 5 dB, with 10 to 20 ms fades, before any processing runs. On a thirty minute interview this takes me under five minutes and does the job no plugin does well.
  3. Insert a dynamic EQ band at the frequency you found, Q around 4, threshold set so it engages only on sibilants, maximum reduction 4 dB.
  4. Compress after that, not before, 3:1 with an attack around 10 ms, aiming for 4 to 6 dB of gain reduction on peaks.
  5. Only if something still cuts through, add a split band de-esser at 2 dB of reduction as a safety net rather than a fix.
  6. Check on earbuds. Monitors flatter sibilance, earbuds expose it, and most of your audience is listening on the second one.

Listen to the processing in isolation if your plugin offers a delta or solo mode. What you want to hear is a faint ticking of S sounds and nothing else. If you can hear words and vowel tone in there, the band is too wide, and you are thinning the voice to catch a consonant.

The mistake I keep making

Treating two voices with one setting. A host and a guest almost never peak at the same frequency, and running a shared insert on a stereo mix means the louder voice sets the threshold for both. Separate tracks, separate settings, and the de-esser never goes on the master. I broke that rule once on a four person panel to save time and spent the next hour undoing a mix where everyone sounded slightly muffled and nobody sounded fixed. Keeping each speaker on their own track also keeps the session tidy enough to reopen next season.

FAQ

Can I remove sibilance after mastering?

Partly, but badly. Once the file is limited and loudness normalised, the S sounds have been pushed closer to everything else and a de-esser has to work harder to separate them. Go back to the recorded track if you still have it.

Does recording at a lower level help?

It helps your peaks, not your sibilance, because the ratio between the S and the vowel stays the same. Set a sensible level, keep peaks near -12 dBFS, and solve the hiss with position instead.

Is there a microphone that just avoids the problem?

No microphone removes it, but a flatter one makes it manageable. Broadcast dynamics are kinder to hard S sounds than bright condensers, at the cost of needing more clean gain from your interface.

Why does my voice sound fine in the room and hiss in the recording?

Your own voice reaches you mostly through bone conduction and from behind the sibilant jet, so you have never once heard what the microphone hears. That gap surprises nearly every first time guest, and it is worth warning them before they listen back.

TB
Tariq Bellweather

Tariq learned audio at a community radio station and now produces interview podcasts. He writes about microphones, rooms and editing speech so that it sounds natural, and he has strong opinions about loudness.

More posts by Tariq

More in Podcasting

01doorganico 084goad 3hot8 4465456 after5internprogram ahandfulofblessings ahmadpc ahomcybersecurity allodialresources antiquesandperiodcollectibles aotc asianworldkashmir beedlectrics bloggertemplatefree bluefeatheragency bodyweightsolution campaign08blog carfixadvisor catstreetcollective cbstvsports cnfanspurchasing committeeof500yearsofdignityandresistance covertopsdjs cs-kaiyuangame cuestacomunicaciontotal cxoutlet dancebetgame deskgearhub destinyvillastay detectmagazine di33 diamondsteelbuildings donmo dyanadropship emarsysfinance emimattl escapechan eskitarih evautoplay everythingweddings-orlando farmzone feedviewerapp foxstudiosscratch gamevui123 gleasongrade greetingsday hacklabasu haskell iw-kaiyun jagoanproperti jilidskripsi joltbusinessfinance juegodepokergratis juegospucca kaeru-festival latestmehndidesigns lestanzediottaviano litemagazines localcommunityhero lustrsazan mariashebets menswheels newwebstechs1 nikkit ootheday oxigenvpn plurfitness prefetial probudgetcarpetcleaning psychicmrsmartin pyaten qvantor rejoignezlogis rexzbot riselayouts ruchin rwxqsgd5 sakaltime shibiaowang shop-bot skrtips spotlesshaven strcoding tetapatrium topselller uneplanteartificielle vipertechcr webtech1223411 wo-kygaming yamaha-mekar yeknava yoybuysheets zenunwa yamaha-mekar jilidskripsi ankertoto ankertoto ankertoto