Audiobook - Small Sample - Welsh voice

Remember you asked.

The background noise is suspiciously low. Nobody records with a nearly -80dB noise floor. Right at “Chapter Forty-One” the voice is hollow-tube honky like it was excessively noise-reduced or recorded in a tiny room… or both.

Make up a test clip for us, and this time don’t “help” it. No effects or filters.

You don’t have to make a production out of gasping for air. I’m going to take that instruction out.


As of right this second, you should see the new clip instructions.


IMO it lacks treble, i.e. slightly muffled.
In order to boost-treble without making sibilance too strong, you need to de-ess* …

before-after , EQ & De_ess.gif

If you are using Audacity on Windows a free de-esser plugin is “SpitFish” …

And if you’re not…


It might be good to wait on quality processing until we get the apparent noise reduction straightened out.

The show might not be quite so smashed and muffled after that.


Paul-L’s De-esser is a Nyquist-plugin so I think it should work in Audacity running on Non-Windows Operating Systems. [ But Paul-L’s De-esser is more complicated to operate than SpitFish ].

Thanks for all the feedback. I’ll get back to you once I speak with the author.

Thanks again.

Good afternoon

OK, I have spent time with the reader and recorded this short sample. No frills, not fixes

Please have a review and advised on the outcome

Many thanks

I need to wait a bit before listening critically. If we do the patching, we can frequently do a better job with less processing. It’s also possible that it’s OK almost as it is and you’re trying too hard.

Nose Reduction recipes get tossed around like walnut-fudge recipes and sometimes you just don’t need that.

As we go.


I applied Steve’s rumble filter just as the clip was supplied to me and I got the attached analysis. Note the last sentence.

Being professionally unable to stop messing with it. I used the very short silent segment at about 8 seconds as Noise Reduction Profile. Then I noise reduced the whole clip by 6,6,6. The Noise Reduction of the Beast. That gives an almost imperceptible suppression of the background hiss, but leaves the human sounds like breathing and putting the coffee cup down. Tea cup? I couldn’t tell.

That’s the attached clip. I would submit that just as it is.

If that’s acceptable, I can publish exact details how I did it.

I see you already know the Normalize to 3.2 trick. Nobody produces peak sound values at -3.2dB straight out of the microphone.


Screen Shot 2015-05-27 at 18.05.14.png

Quick note. You didn’t tell us anything about the computer or Audacity. These advanced tricks work with Audacity 2.1.0.



Many thanks for the feedback, the sound with the 6,6,6, reduction is great and as as you say just enough to remove the hiss.
I’m very happy with the outcome please do drop me a note on how you achieved it all. I have attempted to repair the sample based on your email, not sure if there was other stuff.

  1. samples small section at 8s
  2. Ran noise reduction with 6,6,6 on whole clip
  3. ran normalize to 3.0

Sounds great

Next thing is am I going to have to start this whole book over again, 42 chapters all around 30mins. There is a lot of lip sound that need to be taken out but i expect that the price i pay.
I have made copies at various stages so may not be sooooo bad.

RE Software: I’m on the latest 2.1.0 and I’m windows 7

Thanks again in advance.


Question: there is a lot of talk about Steve’s Rumble Filter, how can I get this I cant find that description is the search?


Pretty sure Koz is referring to the attached equalization curve. I don’t think there is an official download location for stuff like this (although perhaps there should be.)

After you download the file, in Audacity choose Effect->Equalization, then “Save/Manage Curves”. From that window choose “Import” and import the downloaded file. It will then show up as one of the presets in the equalization menu.
LFRolloffForSpeach.XML (453 Bytes)

I’ve posted it here so that it will be easier to find in future: Submit your EQ Curves - #139 by steve

The “official” download location is this page on the wiki, though there are not many XML files there yet: Missing features - Audacity Support

Thanks folks, I’ll give it a go.

My i please ask, how did the NYQUIST report get generated showing the stats for the recording?


If you do an Analysis > Plot Spectrum on the clip, you will see an enormous amount of energy below 100 Hz and indeed a great pile of it below hearing, 20 Hz. That can screw up the accceptance noise reading. LF_rolloff strips all that away.

ACX Analyze
That one’s a flynwill recent product. I can post that when I get to a real computer. It’s still in development (days old) although it’s getting harder to tell the results of that versus the classic method.


There is a lot of lip sound that need to be taken out

Did anyone actually complain about that or are you being obsessive? Anything you do to the work has to be done to the whole work and works thereafter. So think about that when you’re piling on extra chores and tasks that will haunt you later. This is the same thing that haunts people who need a long laundry list of filters and effects just to rescue their performance enough to make ACX compliance. This is even before they filter their voice for pleasant sound.

ACX is very clear they would just as soon you didn’t do that. I use an example of you telling a story to someone in a quiet room over cups of hot tea. Smacks and all.

As far as multiple reading, a recent poster is on her second and third reading of some of her material. They started out stiff and awkward, but the latest efforts compare very well to other’s publications. Listen to your work as compared to other shows that you like. But don’t be obsessive about copying someone else’s style, either. Actor’s get hired in order to contribute their own style.


Thanks for the advice, i must admit I may have been a bit over protective of the lip smacking.

Just one other thought, as the other files may be below -60 does this mean I will need to redo the editing? Is there a way of adding the room tone back into the gaps in the recordings?