How to clean up phone audio using Audacity?
January 22, 2007 11:58 AM   Subscribe

I've got a phone conversation that's going to become a podcast. How can I use Audacity to make it sound its best?
posted by jjg to Computers & Internet (10 answers total) 6 users marked this as a favorite
Normalizing is probably a good place to start (evening out the distribution of volume throughout recording). Hiss removal at the highend, but don't overdo it. If it's a normal phone recording there is a lack of bass and a lack of treble, you cannot re-establish these with an equalizer. They were never recorded. Try instead to get the recording to be comfortable to listen to. Every one in the conversation should be at a similar volume and sound like they were recorded using the same setup. If the various people sound very different, you might want to master a recording for each voice and the edit them back together in the end.
posted by doctor_negative at 12:15 PM on January 22, 2007

Audacity is not the best tool to make this work, first off. If you are going to use it, I would just up the amplitude to baseline -12 or -13 and then pull out all the clicks and static washes, edit out loose empty spaces but leave enough to allow it to feel like a natural conversation between two (or more) people.

Before you edit, get it converted to mono, 16 bit, 44.1 kbps, which is the broadcast standard.
posted by parmanparman at 12:26 PM on January 22, 2007

another thing to do, if the audio is really not great is to put a low-volume music underlay in the track.
posted by parmanparman at 12:27 PM on January 22, 2007

Before you open it in Audacity, put it through The Levelator first. Brilliant tool ... and free.
posted by grabbingsand at 12:51 PM on January 22, 2007 [1 favorite]

Yeah, Audacity is a good tool for some stuff, but to really do this, it's not a great tool. The Levelator should help, though it isn't perfect either. Then I'd use the graphic EQ, and try to find where the hiss and other noise is worse and reduce it.
posted by YoungAmerican at 12:55 PM on January 22, 2007

I don't know if Audacity has it, but please for the love of all that is good, run your sound through some dynamics compression before posting it on the internets. The lack of proper dynamics compression is my single biggest gripe about most podcasts, and I have a lot of gripes.

Dynamics compression will basically even out the volume levels of the voices. If you have for example one person speaking really softly and another really loudly, compression will fix that so they both sound about the same, and normalization won't.

This makes it a LOT easier to listen to your audio in non-pristine environments. I *wish* that my TV, my iPod, etc, all had dynamics compression on their outputs, but alas they don't, and so I have to keep riding their volume controls like maniac to avoid things being too loud for my comfort or too quiet to hear them.
posted by mattly at 1:15 PM on January 22, 2007

The biggest drawback in Audacity is that you can't adjust effects while you are in playback mode; you have to apply an effect over a time range, and then listen to your clip to evaluate the results. I often use Audacity at the beginning and ending of a project, but I add all my cool effects using other programs.

That being said, I would probably start by using the equalizer to eliminate the high end; anything you picked up on the phone over 5 kHz is probably noise. Similarly, you can probably roll off anything on the low end below 500 Hz, but this will probably be much less noticeable.

If the volume is inconsistent, I'd try my luck at some compression. Compression will level the volume over the selected area at the expense of dynamics - for this kind of thing I imagine that you should be able to get away with quite a bit of compression. However, I would be careful particularly if there is any background noise - if you've ever watched a war movie on DVD and struggled to hear the voices over the explosions even though the voices were perfectly audible during the quiet scenes - well, that's compression doing a bad thing.

EQ and compression are going to be pretty much trial and error, keep playing around until it sounds good to you. A spectrum analzyer can also be very helpful in finding hiss and noise, you can then target and EQ out the noisier frequencies. (Sorry, can't remember offhand if Audacity includes a spectrum analyzer.)

Finally, I'd normalize the track to -3 dB. This will adjust the overall volume level of the track, leaving you with ample headroom. I'd do this last to compensate for any level loss due to EQ and compresssion.
posted by malocchio at 1:17 PM on January 22, 2007

What you need to do to the file depends completely on what it sounds like. If the audio sounds reasonably decent, you can probably do some noise removal and compression and be fine.

I've never used Levelator, but I don't see why you'd need to. The leveller in Audacity works fine - it compresses the audio so that the volume stays consistent.

The noise removal plugin is pretty good, too. Select a segment of audio that's pure noise, then you bring up the noise remover and tell it to "get noise profile." Then select all the audio you want to clean up and run the noise remover on it. Experiment with more or less noise removal until it sounds best.

I don't think there's any point in messing with EQ, and the EQ in Audacity is crappy anyway. But if you were going to EQ, a small boost around 3 or 4kHz can be helpful for making speech intelligible, and you could high-pass around 150 or so to get rid of rumbling or plosive sounds.
posted by ludwig_van at 4:08 PM on January 22, 2007

I've recorded phone calls before and the main problem is 60Hz power interference on the low end and hiss/crosstalk on the high end.

You will want to play with the EQ to rolloff the low frequency stuff first, then try rolling off the high frequency stuff next. Remember that telephone only pass 300 Hz to 3100 Hz frequency band, so you can get rid of most everythign else and it shouldn't degrade the quality.

Do the band-pass EQ filter first, then do any normalization. You don't want to normalize when you have spurious noise out of band.
posted by Argyle at 4:56 PM on January 22, 2007

Wow, I guess I didn't realize telephones only did 300-3.1k. In that case, yeah, you should EQ out everything else.
posted by ludwig_van at 5:01 PM on January 22, 2007

« Older can you recommend a home security / alarm company?   |   Dying from cancer... Newer »
This thread is closed to new comments.