Reducing booming

Hi, Windows 10 Pro and Audacity 2.2.2 (I have a lot of chains and plugins that I don’t want interfered with if I upgrade.)

I see from other posts that reducing echo or reverb after the fact is rarely doable. I would appreciate your input on whether or not there are any improvements possible for my current project. I am recording in a room with hardwood floors and wood panelling, and can’t change rooms because this is part of a video and need this location.

Here are two excerpts, the first with a Focusrite Scarlett 2i2 mic about a foot away from the speaker, the second with a Rode NTG3 shotgun mic about two feet away.


Both have one pass of noise reduction applied, and the following equalization, previously provided by Koz:
deboomy.png
They are RMS normalized, and soft limited.

Is there anything that can be done to improve either of these?

Focusrite Scarlett 2i2 mic

The Scarlett 2i2 is an interface, not a microphone.

Is there anything that can be done to improve either of these?

You could try “looping.” That’s what the movies do if they run into a particularly difficult sound problem.

Set your microphone up in a good, quiet, echo-free room. Play a short section of the video repeatedly (loop it) while you record clean, perfect dialog on a separate recorder. The goal is to match your voice to the video. When you get a close enough match, loop the next section of video. Eventually, you will have enough clean, perfect dialog to replace the damaged sound through careful editing.

I am recording in a room with hardwood floors and wood panelling, and can’t change rooms

Thus practically guaranteeing bad sound. It is the Producer’s job to arrange shoots so both picture and sound work well. I would get a new Producer.

Koz

It is the Producer’s job to arrange shoots so both picture and sound work well. I would get a new Producer.

I sent myself a strongly worded email to this effect.

I then replied that I was a little surprised to receive my email, since, as I knew very well, these were requirements I myself had imposed.

I replied that I was not interested in excuses, and required results, and expected that I would find a way to achieve what I had requested.

I responded that if this was the way I was going to treat myself, maybe I should just find another producer.

I replied that if that was they way I was going to respond, then maybe that was exactly what I should do.

“Fine!” I replied.

“Fine!” I responded.

A little while later I sent an update that perhaps we had let emotions get the better of us, and on some quiet reflection, I could understand where I was coming from, and in the interests of the best outcome for the project, if I was willing to talk about it, I would be happy to discuss further.

Thankful to receive the message, I immediately replied that I felt the same way, and really valued the contributions I had made, and would not have been able to make the progress I had so far without me, and really wanted to keep this valuable team together, so perhaps we could have a coffee together?

I had a coffee with myself, and agreed that we had always been a fantastic team, and we should never let anyone come between us again, and reached a mutual agreement that it was really all Koz’s fault.

Giving myself an embarrassed hug, I resumed work on the project, thankful we had managed to work all the issues out, and resolved to never let anything like this come between us ever again!

That made me laugh :smiley:

I totally sympathize with both of “you”, and pleased to hear that you have resolved your issues.

I resumed work on the project, thankful we had managed to work all the issues out, and resolved to never let anything like this come between us ever again!

Thus neatly side-stepping a fix for the actual problem—bad studio acoustics.

Echoes are your own voice arriving at the microphone multiple times: Once direct and then multiple times bounced from the walls, ceiling, and floor. So the job is removing your voice from itself. That’s why nobody is racing to the rescue. This is not trivial.

You might well say the echoes are always lower volume than the voice. True, and that’s how Noise Gates work.

Anything lower volume than what the plugin perceives is normal speech is suppressed. That seems like a gift from the angels, but it comes with ancillary problems. You have to announce perfectly. No more expressive or theatrical volume changes. One volume change in the wrong place and your voice—or parts of your voice—will vanish.

Also, they can damage normal speech by accident. A wrong setting can chop off leading and trailing word sounds.


I can provide several instances where someone produced perfect visuals and destroyed the whole thing by sounding like they’re announcing in a bathroom.

Q1, Q2

You’re juggling distortions. Q2 does have fewer echoes and that would be unconditionally recommended, but it has Essing. Your SS sounds could cut wood and scare the cat. I use Trebor’s DeSibilator rather than the more common DeEssers.

Select your performance and find the RMS (loudness) value. Analyze > Contrast > Measure Selection (write it down) > Close.

Plug that into the DeSibilator “Threshold” setting. > OK.

Koz

Great advice. We both thank you.

The free version of Couture (expander plugin) can attenuate some of the room reverb…

You might ask the Producer when is constant haggling and messing with marginal sound worth getting a proper studio and ditching the polished hardwood floor and rich wooden paneling.

Sound without the picture is a radio show. Picture without the sound is a rehearsal.

That’s not to say there are no ways to force this to work. Make the walls non-parallel.

We had a studio at work that appeared to be a plain room with a little industrial carpeting on the floor and I’m not even sure if it had acoustical ceiling tiles. It was a terrific studio for sound and I pushed several productions through there. It was used for musical rehearsals. If you have a big enough company, you always have enough people for a band.

Email @etc. “Anybody play bass guitar?”
[Johnny-In-Graphics] “I do.”

The trick was the walls weren’t straight/square/parallel and neither was the ceiling. On purpose. Echoes died a very rapid death in that room. It sounded dead or at least non-bathroom without needing to have fuzzy walls and expensive panels.

On the other hand, I had an office with surgically square, bare walls. I could clap and it was still bouncing back and forth when I got back from lunch.

Clap clap clap clap

The exact worst room in the building to shoot sound.

Much like yours.

Koz

The free version of Couture (expander plugin) can attenuate some of the room reverb…

I’m trying to set it up properly, and as usual my old ear is not good enough to decide myself. Is the best thing to match the settings in the picture you linked to? That is, from left to right:

  • Detect=Sybil
  • Clip=20
  • Transient=x1
  • Speed at 3 o’clock
  • Front/Back at 3 o’clock
  • Output at about 83%

Or should I use other settings?

Also, instead of a DB slider, the couture I just downloaded has a couple extra options by the output knob, default Dry=-infinity and Wet=0dB, as below. Are these the best?


If buying the paid version would be better, $19 is not a problem, just let me know.

Thanks.

You can tell visually when it’s right from the red line on the Couture display: around 6dB reduction when you are not speaking is about the maximum attenuation you can get away with without it sounding choppy …




-infinity dry is the same as the old version of the plugin, which did not have those sliders.
(Any dry signal is going to have the room reverb you are trying to reduce).


All the expander controls work on the free version: no advantage in paying for the full version.

You can tell visually when it’s right from the red line on the Couture display: around 6dB reduction when you are not speaking is about the maximum attenuation you can get away with without it sounding choppy …

So I want the redline to be just at, or below the black line?

Any reason you have it set to Sybil instead of Human?

To make sure I am not missing something, would it be possible for you to upload the preset you are using in your example?

Any reason you didn’t opt for the shotgun with the DeSibilator to get rid of the gritty SS sounds?

Koz

Any reason you didn’t opt for the shotgun with the DeSibilator to get rid of the gritty SS sounds?

Definitely doing that too.

When the red line corresponds with the black line it’s not doing anything.
when the red line goes below the black it’s attenuating the output.

Supposedly Sybil removes more sibilance than Human.

3-o’clock-ish.zip (265 Bytes)
Bear in mind my example also had bass-cut equalization (EQ) & desibilator.

Here is an excerpt with Couture first (using Trebor’s preset) and then Desibilator, then the opposite order Desibilator and then Couture.

I can tell a very very slight difference, but am not sure there is a quality preference. What do you think?

Anything else you would recommend to improve this?

IMO you’ve overdone the de-essing a little : “once” sounds like oneth.
& you could remove more bass, (<200Hz), as most of the room-reverb lives there.

Dare I mention the mic clicking ?

IMO you’ve overdone the de-essing a little : “once” sounds like oneth.

I hear that. I am using your (Trebor’s) DeSibilator at

with Koz’z direction to “Select your performance and find the RMS (loudness) value. Analyze > Contrast > Measure Selection (write it down) > Close. Plug that into the DeSibilator “Threshold” setting. > OK.”

What should I do to dial this back?

you could remove more bass, (<200Hz), as most of the room-reverb lives there.

I’ve already applied Koz’s “de-boomy” equalization from https://forum.audacityteam.org/download/file.php?id=31565. To take out more from below 200Hz, do I instead just make the graph fall off for everything from 200 Hz down? Or is there a preset that will do this?

Dare I mention the mic clicking ?

Oh yes, please dare. These videos will hopefully last on Udemy for decades, and this is my one chance to get them right. What do you recommend?

Have we ever experienced your raw voice without the basket of corrections? If you apply enough filters, effects, and corrections, you start putting out self-generated fires. I like DeSibilator because it doesn’t turn SS sibilance into FF damage. I have no idea how you got that problem to come back.

How about you post a ten second voice test with no corrections at all.

https://www.kozco.com/tech/audacity/TestClip/Record_A_Clip.html

Koz

Many thanks - here you go, with raw voice from the NTG3 through a Scarlett Studio 2i2:

The sample we were working with then had the following processing:

  • noise reduction
  • equalization 100 Hz rumble
  • equalization de-boomy setting shown in original post
  • change pitch -5%
  • RMS normalize -18
  • soft limit -3 dB

Then:

  • couture with Trebor’s preset
  • desibilator with a threshold from the analyze measurement of -18.79 (both foreground and background were giving me the same measurement).

Does this cast light?

RMS is a first-approximation for the desibilator threshold,
if there is lisping with the threshold at RMS value,
try raising the threshold from the RMS value in 1dB steps until the lisping goes away.
e.g. if the RMS is -20dB, but preview sounds lispy, try -19dB, -18dB, etc


I would use a real-time equalizer plugin, e.g.
so I could hear the effect of adjusting the bass as the track played.
IMO attenuate everything below 200Hz, (not just a dip between 100-200),
as some people will be able to hear <100Hz, depending on the device they are listening on.


It sounds like clicking is occurring in the mic, (rather than in your mouth).
Paul-L’s is the only free de-clicker plugin I know of.