hpr-knowledge-base/hpr_transcripts/hpr3004.txt

Episode: 3004
Title: HPR3004: Fixing simple audio problems with Audacity
Source: https://hub.hackerpublicradio.org/ccdn.php?filename=/eps/hpr3004/hpr3004.mp3
Transcribed: 2025-10-24 14:53:59

---

This is Hacker Public Radio episode 2004 for Thursday, 6 February 2020.
Today's show is entitled Fixing Simple Audio Problems with Audacity.
It is hosted by Dave Morris
and is about 13 minutes long
and carries an explicit flag. The summer is
sharing a few experiences with Audacity that may be helpful to others.
This episode of HPR is brought to you by archive.org.
Support universal access to all knowledge
by heading over to archive.org forward slash donate.
Music
Hello everybody, this is Dave Morris
and I'm welcoming you to Hacker Public Radio.
Today I want to do a hopefully quite brief show
talking about using Audacity to solve a few audio problems.
Now I'm no expert on the use of Audacity.
Have a bit of experience and have managed to fix a few things.
But I thought in sharing a few recent experiences
might be helpful to some people.
Maybe get a conversation going and get some more contributions
to this subject from people who prefer to know a lot more than I do.
So what happened was I recorded the audio for a show I did with Mr. X in 2019.
Number 2972 it was.
And we were sat in my car because we couldn't finally wear quite
and it was very cold so you didn't want to be outside.
And I used my Zoom H2N recorder when I'm using now actually.
And I put it on a little tripod on the dashboard of my car.
But something about this setup caused to me anyway
the resulting audio to be very sort of boomy and echoey
and quite a lot of sort of bassiness in it.
And I was keen to reduce that, make it a little less unpleasant to listen to.
Now one of the things that I do with pretty much all of my audio
partly because when I'm speaking, when I'm doing these types of things
I tend to have a script so much more.
I tend to be looking at the notes that I've prepared already
and I tend to be sort of riffing on them.
And along the way I sort of hesitate and leave gaps.
So I thought I usually run the truncate silence effect in audacity
to clean that stuff up.
But I have not been doing this very well I feel
because listening back some of the results
you can hear that if you speak as I do,
where you sort of try to get very quiet towards the end of a sentence.
The truncate silence just chops that off.
And if you don't start a word very loudly,
then it can also do the same.
And you do hear on other stuff as well.
And it is a common issue with the silence truncation I think.
Anyway, I thought in this particular case
I would try and get truncate silence to work better
or learn how to use it better.
So I put a couple of requests in the notes here
if I didn't do a good job explaining all this stuff in this show,
then let me know and tell me what I should have done.
And if you have the expertise, then how about doing an HBR show
to talk about how you deal with common audio problems?
For example, I've had things where there's a mains hum
probably because where I record and the proximity of,
I'm in my sort of dining kitchen area
and there's things like fridge freezes too, actually, quite close by.
So I think sometimes mains hum gets into stuff.
So how would you remove that?
And do you use compression and normalization to clean things up?
That type of thing would be amazing to know more about, I feel.
So what did I do with the audio of this episode 2.9?
7.2
Well the first thing I did, and I always do, is to run the noise reduction effect.
And that's because most of what I produce has some background noise
and I need to chop that out.
What I have to do is to select the effect
and sample a piece of the audio to get what they call a noise profile.
What I do usually is to switch the recorder on,
let it run for a little while, a few seconds.
So I've got a sample of noise to do this with.
And then having taught the effect what the noise profile looks like,
then I select the whole thing, tune settings a bit and run it on the whole thing.
And you should see the silences that the amount of noise in the silences
disappearing somewhat.
So that's the thing I find really, really useful to do.
I'm not going to go in a lot of detail about how you do this
because if you go and look at the audacity manual and I've linked it,
it shows diagrams and explains it really, really well,
much better than I can do.
The second thing I did was I ran a high pass filter on the audio.
Now this is a way of reducing low frequency noise.
And as I said before, I felt that there was the hollow nature of the original audio
was perhaps due to the audio frequencies that were knocking about inside my car.
It wasn't damped properly or something.
I really don't fully understand somebody who's a proper audio engineer
might be able to explain why it would be like that.
I certainly heard other people recording their cars and it doesn't sound too bad.
Anyway, I thought, how about if I remove, if I try a high pass filter?
And what this does, it passes the frequencies above a certain threshold that you set.
And attenuate frequencies below that cutoff frequency.
So I set the cutoff frequency to 500 hertz and there's a roll-off setting,
which is the decibels per octave, two six decibels.
I had tried what the effect offered me as the default,
which was a thousand hertz in the first instance, but it was awful.
So I think it actually removed much, much of it.
It doesn't really sound like a human speech at all.
So 500 seemed to be a good compromise.
And that seemed to do something useful.
You could hear, to me, anyway, you could hear the reduction of some of the frequencies.
It seemed to me to take away that sort of weird boominess.
Now, because that high pass filter introduced the overall volume,
I then applied an amplification pass.
And this is quite well documented and it's fairly easy to do.
There's a value in the amplification box when you fire up this effect.
And if you follow that, it will produce a new peak amplitude of zero decibels,
which is sort of normal level, I guess you'd say.
And in the case where I did this, the value was 1.106 decibels.
So I just used that.
Because you can do these things, you can run the effects,
listen to what the result is, and then click undo, you can experiment.
Especially if you make sure you've saved the work so far before you do it.
I might not have amplified it quite enough, because the end result sounded moderately quiet
when listening to it on an HPR.
The waddley, it sounded fine when played through audacity.
But obviously the settings are different in the two situations.
But anyway, it bumped things up a little bit more so it didn't sound quite so faint
after the work already done on the sound.
Then came the silence truncation, which I tend to do at the end of everything.
This is the sort of order I normally use if I'm doing this sort of stuff,
which again I'm open to to comment on.
It doesn't seem like I'm doing this right.
So as I've said in the notes is an art to using silence truncation.
And I said in the preamble that the effects it has,
if not used correctly, is that it chops the beginnings and ends off words.
When I started doing this on this particular audio, I did it the way I normally do it.
And there were some really unpleasant truncations applied to it.
So this may be one of going experiment with how to do a better job.
The truncate silence effect is documented really well.
So I've referred to that.
What I did was I set the threshold to minus 25 decibels.
I tried using minus 20, but that caused some of the quieter starts of words to get truncated.
So the decibel scale is logarithmic and zero is sort of the normal, the mean, I guess.
And so it'll probably explain it better in the audacity manual.
Anyway, I use minus 25 decibels.
Then I went for a duration of half a second, 0.5 seconds.
I used the compress excess silence option and shows compressed to 50%.
Now you'll see this in the documentation.
Now this is the first time I'd used this combination of settings.
Previously I'd chosen the truncate detected silence option and had not been using the best threshold value to do it.
So I think that the end result is a lot better.
It doesn't, it's not quite a brutal with removal.
So if you've gone a long piece of audio with silences in it,
it doesn't reduce it quite as much as the other one would.
However, the end result sounds a lot nicer.
The end result sounded pretty good.
And I'm hoping that this will help people too.
They're not doing something like this to have a go at it if the mood takes them.
So what I'm going to do is to play the same chunk of audio from the show I've been talking about.
And with the different effects.
So the first one is the original audio as it came off the recorder.
It's got noise in it.
I think it sounds boomy, but more I look at this the less I'm convinced.
However, bursting anyway.
So here it is.
Hi everybody, this is Hacker Public Radio.
Welcome to our show.
So the next one is after noise reduction.
Doing the same sort of things I mentioned before.
I haven't taken the noise from this piece of audio itself,
but from the preamble in front of it, which I haven't included.
So if you can see the waveform, you'd see that it's a lot cleaner.
Hi everybody.
This is Hacker Public Radio. Welcome to our show.
Next is the high pass filtering.
And you can hear it. It's got a lot more...
Sounds like an old-fashioned telephone in some respects.
The number of frequencies have been reduced. You can tell that.
And from the waveform, you can see that stuff has been reduced.
The peaks have been reduced in height, which is why.
Hi everybody.
This is Hacker Public Radio. Welcome to our show.
The next one, I have thought that the volume was too low,
so I've amplified things.
And this is the result.
The waveform looks a lot higher than the previous one.
Hi everybody.
This is Hacker Public Radio. Welcome to our show.
And the last one has had some silence truncation.
There wasn't much silence in this sample.
There were two of us talking, so whenever there was a gap, somebody else spoke.
So there wasn't a lot to remove.
However, I wanted to demonstrate that it's possible to do silence truncation
without nibbling away at least starts and ends words.
So that's the end product.
Hi everybody. This is Hacker Public Radio. Welcome to our show.
So hopefully that demonstration helps a wee bit.
I hope you found it entertaining, even though you had the same clip over and over and over.
I got bored, so you probably got a little bit.
Okay, that's it then. Bye bye.
You've been listening to Hacker Public Radio at Hacker Public Radio.org.
We are a community podcast network that releases shows every weekday, Monday through Friday.
Today's show, like all our shows, was contributed by an HP R listener like yourself.
If you ever thought of recording a podcast, then click on our contributing to find out how easy it really is.
Hacker Public Radio was founded by the digital dog pound and the Infonomicon Computer Club
and is part of the binary revolution at binrev.com.
If you have comments on today's show, please email the host directly.
Leave a comment on the website or record a follow-up episode yourself.
Unless otherwise stated, today's show is released under Creative Commons,
Attribution, Share a Life, 3.0 license.