How to fix harsh s sounds in your podcast
- Step 1Drop your edited episode in — After your cut is done, drop the episode (or an individual track) onto the de-esser. MP3, WAV, FLAC, M4A, OGG, Opus, or a video file are all accepted. It processes one file per run.
- Step 2Spot-check the worst sibilant moment — Scrub to a line with lots of esses ('she sells seashells' energy) and listen on the headphones you'll publish-monitor with. That tells you how aggressive the Intensity needs to be.
- Step 3Dial the Intensity for your mic — Set the single Intensity (0-1) slider. Start at the
0.6default (-7.2 dB). A dynamic-mic voice often needs less (0.4-0.5); a bright condenser guest may need0.7-0.8. Step is0.05. - Step 4Process and download the de-essed file — FFmpeg WebAssembly applies the notch and re-encodes to the same format you dropped in. The de-essed episode downloads to your machine — nothing was uploaded.
- Step 5Standardise the setting for the show — Note the Intensity that worked. Because the curve is fixed, the same number gives identical de-essing on the next episode recorded with the same mic — no re-tuning, which keeps a series sounding consistent.
- Step 6Finish the chain in publish order — De-ess, then level dynamics with speech-leveler, then hit your platform LUFS with loudness-normalizer (-16 for Apple Podcasts, -14 for Spotify). Or do the whole thing at once with podcast-master.
Intensity by microphone type
Starting Intensity values for common podcast mic setups. The cut depth is exactly intensity x -12 dB at the 7-9 kHz trough. Always A/B and adjust to taste — these are starting points, not rules.
| Mic / setup | Suggested Intensity | Cut depth | Why |
|---|---|---|---|
| Dynamic (SM7B, PodMic) | 0.4-0.5 | -4.8 to -6 dB | Dynamics are naturally darker; usually need only a light shave |
| Large-diaphragm condenser | 0.6 (default) | -7.2 dB | Bright and revealing — the default notch suits most |
| Bright/cheap condenser or USB mic | 0.7-0.8 | -8.4 to -9.6 dB | Exaggerated top end pushes more energy into 7-9 kHz |
| Lavalier / clip-on | 0.5-0.6 | -6 to -7.2 dB | Variable; depends heavily on placement near the mouth |
| Already-EQ'd master with top boost | 0.6-0.7 | -7.2 to -8.4 dB | Your presence boost amplified esses; de-ess after the boost |
Where the de-esser sits in podcast post
Recommended order of operations for a podcast episode. Each step is a separate JAD audio tool; the de-esser handles only the sibilance step.
| Order | Step | Tool | Why this order |
|---|---|---|---|
| 1 | Remove background noise | ai-noise-reducer | Clean the signal before any tonal work |
| 2 | Tonal EQ (high-pass, presence) | voice-eq | Shape the voice before taming its harshness |
| 3 | De-ess sibilance | This tool | Cut the esses the EQ presence boost amplified |
| 4 | Even out dynamics | speech-leveler | Level loud/quiet words after the tone is set |
| 5 | Normalize loudness | loudness-normalizer | Hit the platform LUFS target last |
Cookbook
Real podcast scenarios with the Intensity to start at and what to listen for. The single slider keeps each recipe short.
Solo host on a condenser — the default just works
The most common podcast case. A large-diaphragm condenser close to the mouth makes esses sharp; the default Intensity tames them without losing the broadcast brightness listeners expect.
Setup : solo host, condenser mic, edited MP3 Control: Intensity = 0.6 (-7.2 dB notch) Result : Esses sit back; voice still bright and clear Next : loudness-normalizer -> -16 LUFS for Apple Podcasts
A guest's USB mic is way brighter than yours
Remote interviews mix mics. If a guest's cheaper USB mic is harsh, set the Intensity to the guest's needs — your darker dynamic mic has little 7-9 kHz energy, so it is barely affected by the same setting on a combined track.
Track : combined host + guest WAV
Problem: guest USB mic is piercing; host SM7B is fine
Control: Intensity = 0.75 (-9 dB notch)
Result : Guest smoothed; host nearly untouched (little
energy to cut in the band)Standardising de-essing across a season
Because the curve is fixed, the value you lock in carries over. This is the workflow advantage of a non-dynamic de-esser: zero per-episode tuning.
Episode 1: Intensity = 0.6 sounded right
Episode 2-12: drop file -> set 0.6 -> process
Guarantee: identical de-ess treatment every episode
(same firequalizer curve, same depth)FLAC interview archive stays FLAC
If you archive masters as FLAC, de-essing keeps the format. The output is re-encoded to FLAC, so your lossless archive stays lossless.
Input : ep042-master.flac
Control: Intensity = 0.55
Output : ep042-master-deessed.flac
Note : no format dropdown for the de-esser; output
always matches the input formatOver-cooked it — pull the slider back
A guest complained the de-ess made them sound dull. Because there is no dynamic behaviour to blame, the fix is purely a smaller Intensity.
Feedback: 'I sound lispy / muffled' Was : Intensity = 0.9 (-10.8 dB) Fix : Intensity = 0.55 (-6.6 dB) Result : Sibilance handled without the dullness
Edge cases and what actually happens
Mixed-mic episode, one host fine and one harsh
SupportedThe notch is applied to the whole file, but a dark dynamic mic has little energy in the 7-9 kHz band, so it is barely touched while a bright mic gets the full cut. Set the Intensity for the harsh voice; the other speaker stays essentially as-is.
De-essing after loudness normalization
Wrong orderIf you normalize loudness first and then de-ess, the de-ess cut lowers the level again and you drift off your LUFS target. Always de-ess before the loudness pass so the final measurement is taken on the de-essed audio.
Episode longer than 120 minutes on Pro
RejectedPro caps each file at 120 minutes. A long-form 3-hour episode is rejected on Pro for duration even if under 200 MB. Split it with audio-splitter, or upgrade to Pro-media / Developer for unlimited duration.
Trying to de-ess on the Free plan
Pro requiredThe de-esser is Pro-only. On Free it will not process. Upgrade to Pro for full-episode de-essing (200 MB / 120 min per file).
Wanting per-word dynamic de-essing
Not availableThis tool applies a constant EQ notch, not a level-triggered dynamic de-esser. It does not duck only on sibilant peaks — the cut is always on. For surgical, transparent dynamic de-essing on a single host vocal, a DAW plugin is the right tool; for whole-episode consistency this fixed curve is usually plenty.
Music sting or intro bed in the file
Affects highsIf your episode export includes a music intro, the notch also cuts cymbal/hi-hat energy in that section. It is harmless at low Intensity but can dull a bright music bed. De-ess the voice track before adding music if you want the bed untouched.
Output came out as the input format, not MP3
By designThere is no format selector for the de-esser. Drop a WAV, get a WAV; drop an MP3, get an MP3. Convert afterwards with a tool like wav-to-mp3 if your host needs a specific format.
No noticeable change at low Intensity on a dark voice
ExpectedA naturally dark voice already has little 7-9 kHz energy, so a 0.3 Intensity may do almost nothing audible. That is correct — there is little there to cut. If the voice genuinely isn't harsh, you may not need the de-esser at all.
Plosives (P/B pops) still present after de-essing
Out of scopePlosives are low-frequency bursts, the opposite end of the spectrum from sibilance. The de-esser cannot touch them. Use a high-pass filter via voice-eq or re-record with a pop filter.
Frequently asked questions
Where does de-essing fit in my podcast workflow?
After noise removal and tonal EQ, but before leveling and loudness. The recommended order is: ai-noise-reducer -> voice-eq -> de-esser -> speech-leveler -> loudness-normalizer. Or do the whole chain at once with podcast-master.
Is this a dynamic de-esser like the ones in Audition or Logic?
No. Those duck the sibilant band only when it spikes. This tool applies a fixed EQ notch across 4500-12000 Hz (deepest at 7-9 kHz) for the whole file, depth set by one Intensity slider. It is simpler and perfectly consistent across episodes, but it is always-on rather than level-triggered.
What Intensity should a podcaster start at?
The default 0.6 (a -7.2 dB notch) suits most condenser-mic voices. Dynamic mics often want less (0.4-0.5); bright USB or budget condensers may want 0.7-0.8. Set it once for your mic and reuse it — the fixed curve means the same number behaves identically every time.
Can I de-ess every episode the same way?
Yes, and that consistency is the point. Because the firequalizer curve never changes, the Intensity you settle on gives identical de-essing on every episode recorded with the same mic. No per-episode re-tuning, which keeps a series sounding uniform.
Will my unreleased episode get uploaded?
No. The de-esser runs on FFmpeg WebAssembly in your browser. Pre-release episodes, NDA interviews, and sensitive guest material are processed entirely on your own machine and never sent to a server.
Why is there no preset menu?
The single Intensity slider already spans the range a preset menu would offer: light (~0.4), medium (0.6 default), and strong (0.8-1.0). Set the number directly — there is no separate Mild/Medium/Aggressive dropdown in the de-esser.
How long an episode can I process?
On Pro, up to 120 minutes and 200 MB per file. Most episodes fit comfortably. For longer raw sessions, split with audio-splitter or upgrade to Pro-media / Developer, which remove the duration cap (up to 100 GB).
Does de-essing lower my episode loudness?
A little — it removes high-frequency energy. That is why you de-ess before the loudness pass. Run loudness-normalizer afterwards to hit -16 LUFS (Apple Podcasts) or -14 LUFS (Spotify) on the already-de-essed file.
Can I batch-process all my episodes at once?
Not in one run — the de-esser takes one file at a time. Process episodes sequentially with the same Intensity. The fixed curve guarantees identical treatment, so a batch done one-by-one still sounds consistent.
What output format do I get?
The same format you dropped in. WAV in, WAV out; MP3 in, MP3 out. There is no format selector for the de-esser. If your host requires a specific format, convert after de-essing with the matching converter tool.
The de-esser is not working on my account — why?
It requires Pro or higher. On the Free plan it will not process. Upgrade to Pro for full-episode de-essing; Pro-media and Developer add unlimited duration and 100 GB files.
My voice still sounds harsh but it's not the S sounds — help?
Then de-essing won't fix it; the tool only cuts 4500-12000 Hz. Boxiness or honkiness lives lower — shape it with voice-eq. Hiss or room noise is a job for ai-noise-reducer.
Privacy first
Every JAD Audio tool runs entirely in your browser via FFmpeg (WebAssembly) and RNNoise. Your audio files never leave your device — verified by zero outbound network requests during processing.