How to EQ Your Dialogue from Start to Finish: Editing Step 6
(Hear audio example in blue box)
Sibilance, those harsh, ear-splitting "S" and "T" sounds in words, is almost always a problem in our dialogue recordings.
It creates a huge barrier to a professional sounding voice recording if our listeners are having to fight through a constant barrage of stabbing sibilance to their ears (as we've talked about in other articles).
So how do we use EQ in the editing phase to alleviate this huge problem and create a soothing sound in the upper midrange of our voice recordings?
What Does It Sound Like?
The sound of sibilance emanates from words with "S", "T", "Z", "Sh" or sometimes "Ch" sounds.
They will often jump out of your speakers at a louder volume than the rest of the spoken words.
These sharp or harsh sounds can definitely be distracting, but could even be painful to your listener if they're not controlled.
That's because they occur in and around the frequencies that our human ears are most sensitive to.
- Listen to the audio example (at top of page in blue box) to hear what sibilance sounds like in my voice-over for video project
- Hear sibilance in these words: "start", "sound", "voice" etc.
- See the sibilant frequencies from the same audio spiking in the frequency spectrum below right under 5000 and 10,000 Hz (Hertz):
How Do You Find It?
One of the few redeeming aspects of sibilance is that it isn't hard to find.
Generally, you're going to hear the piercing "S" type sounds very easily as they rear their ugly heads, popping out above the rest of the audio at a higher volume.
As you saw in the image above, it's also pretty easy to identify it on a frequency spectrum analyzer (get yours free here).
Look for the spikes in the upper midrange of your voice recording somewhere between 3000 and 10,000 Hz.
Oftentimes, you'll see it spiking in multiple places up and down the upper midrange frequencies.
For example, the main sibilant problem areas in my voice-over project (see below) were around 3900, 4700, 5600 and 9600 Hz.
Finally, you can also use the same method we used for finding resonances - boosting and sweeping a bell filter through the upper midrange area to pinpoint nasty sibilant spots.
How To Remove It (Taming the Beast)
Now I'll show you one way to remove sibilance using the plug-in I suggested in the removing resonances article: the FabFilter Pro Q3 EQ.
This EQ has a dynamic EQ function that is wonderful for combatting sibilance and is very easy to use.
Plus, it's already optimized, so you don't have to worry about dialing in attack or release settings.
The benefit of dynamic EQ is that it can reduce sibilance more precisely than other methods leaving good top end clarity intact.
It works by only reducing in the area you select and then only when sibilance spikes and exceeds a set threshold.
To remove sibilance with the FabFilter Pro Q3 dynamic EQ:
- start with a bell filter boosted a few dB using a Q of 5
- search around in spiking areas in the upper midrange using the solo listening function
- you should hear the sibilance poking out pretty easily listening to only your selected area (this is the solo listen function)
- also see the sibilant frequencies pop out of the EQ's spectrum as you hear them
- set up a dynamic EQ band in bad sibilant spots by clicking the drop down arrow to the lower right of the small box hovering over your selected area
- select "Make Dynamic" and you'll see the thick yellow curve reducing when sibilance occurs
- adjust your dynamic band for what sounds right by dragging up or down on the area underneath the reduction band
- you can key in exact amounts for the dynamic band by clicking in the red lighted ring area on the "Gain" knob (in floating band control near bottom of display)
- next, narrow the Q down to the specific sibilant area (to around 7) or keep it more broad if there are several areas of sibilance in your band
- my example below had both: spots where I needed to zero in on a very bad sibilant frequency and...
- broader areas where I needed to cover several spiking sibilant frequencies at once
- finally, adjust the amount of reduction until the sibilant spots are smoothed out and less harsh when listening back...
- drag down on the area underneath the dynamic band for more reduction (less sibilance)
- if you start hearing a lisping sound where there was sibilance, you've gone too far - drag back up for less reduction
Here are my exact settings for de-essing my voice-over audio (see image below) and what I was thinking:
Notice the thick yellow curve showing where and how much reduction is happening when the sibilance hits.
You can see the thick yellow curve's cut is deepest on the blue band since it has the most reduction, -8 dB, of all the dynamic bands.
Band 1 (green):
- at 2820.1 Hz
- -3.00 dB (dynamic band reduction)
- 7Q (width of band)
Band 2 (blue):
- at 3879.4 Hz
- -8.00 dB (dynamic band reduction + -2.00 dB EQ cut)
- 8Q (width of band)
Band 3 (red):
- at 4676.5 Hz
- -4.00 dB (dynamic band reduction)
- 5Q (width of band)
Band 4 (yellow):
- at 5565.1 Hz
- -3.00 dB (dynamic band reduction)
- 7Q (width of band)
Band 5 (pink):
- at 9634.8 Hz
- -1.50 dB (dynamic band reduction)
- 5Q (width of band)
I found my main spiking problem area around 4676 Hz (the red band above) and then filled in around it with other bands to address other sibilant spots.
I tended to zero in (narrow the band) and use more dynamic EQ on really bad spots.
See the bands above that are pulled down to -3 dB or more (dB settings on far right in red box; pinch to zoom in on phone).
Most bands ended up needing to be at -3 dB or more since I could hear that the sibilance wasn't being reduced enough (needed more reduction from the dynamic band):
- I set the dynamic EQ at -3 dB by clicking into the lighted area of the "Gain" knob in the floating band control to type in "-3".
I even did a regular EQ reduction added to the dynamic EQ in the worst spot at 3879 Hz (the blue band) to tame a terrible sibilant whistling sound in my voice:
- -2 dB EQ cut together with -8 dB dynamic EQ; Q of 8.
The sibilance in the highest frequencies, around 9634 Hz (the pink band), was the least problematic so I stuck with the default dynamic EQ setting:
- however, I made the Q more broad in the pink band (left it at 5) to cover sibilance in several areas around 9634 Hz.
Let's Review
Although some of this may seem technical, think of these steps as just guidelines as you de-ess your voice recordings.
Remember, to recognize the sound of sibilance, listen for harsh "S" and "T" sounds jumping out in your podcast or sermon audio.
Finding the sibilant area isn't usually too difficult as you listen and look at a frequency spectrum analyzer like Voxengo's Span.
You'll see the sibilant areas spiking in the upper midrange frequencies and hear their piercing quality well above the non-sibilant words.
Finally, although there are other good tools, try using a dynamic EQ since they are well suited to deal with sibilance in a precise way that doesn't destroy the upper midrange clarity of your dialogue.
Try using my framework above to set up your dynamic EQ today!
As always, make sure it sounds better.
Compare the "before" and "after" versions to make sure the de-essing has helped.
Once you get the hang of smoothing out the sibilance, your listeners' ears will thank you forever!
Have YOU used these steps on a recent voice-over for video or audiobook recording? How did they help?
Cheers, Bryan
Your Dialogue Editing Mentor
Transform Your Voice Recordings to Sound Professional
Grab My 15-page Guide Packed With Tips To Easily Improve Your Audio Today!
When you sign up, we'll send emails once every 2 weeks with valuable free content. We promise to keep your info safe.