Can I reduce mp3 file size?

I exported a podcast to mp3 and got a file that was 23 MB. :astonished: My Drupal website objected and told me 5 MB was the limit. So here I am, trying to figure out how I did it before, because a few years ago I produced some podcasts and they came out the other end at about 5 MB? hmmm?

Soooo, any ideas how to compress these things? :question: And do I pay a price with quality?

And do I pay a price with quality?

Of course. Everything MP3 does sacrifices quality. The best you can ever do is minimize it.

The worst thing you can do is make an MP3 from another MP3. The compression damages add up. You can get into trouble pretty easy. If you have a 128 quality MP3 and you cut it into a show and export it 128, the real show quality will come in something around 64.

When you make the MP3 in Audacity 2.1.2, There will be an option panel or a built-in info panel during export where you can change the compression quality. The fuzzy rules are 32 for mono, 64 for stereo are the lowest you can go before everybody notices Something Wrong. I don’t know of a way to predict the filesizes. I just experiment.

If all you’re going to do is cut and simple edits, you should probably use one of the pure MP3 editors rather than Audacity. They don’t have the re-encoding step and their sound damage doesn’t increase. If you need heavy editing and effects, you’re stuck.

We warn people never do production in MP3. Record your podcast and cut it in WAV. Use WAV for the archive storage and only then make the MP3 for posting. If you have to make a correction or change, do it to the WAV, and make a whole new MP3.


A few years ago, streaming had a lot lower bitrate. And podcasts were shorter. So I’d guess that the resulting file would be smaller.

And like Koz explained, it’s all in the quality. With a lower bitrate, the file will be smaller.

My podcast is 25 minutes long, mono, 22050 hz, 32 bit float.

Excuse my ignorance, but I presume bitrate refers to 32 bit float. So I plan to experiment. Should I export to wav and play around with the wav file, then re-export the wav file to mp3? Is that the procedure?

No the 32-bit float is the sample rate. How finely and accurately does the system hack up your analog sound to turn it into digital. 32-float is fantastic overkill. Audio CD quality is 16-bit. So is digital television.

Why did you decide on those sound specifications? They’re really odd.

Bitrate is the number of bits required to transmit your show. As a completely stupid example, suppose you and a friend knew when you said the number “4” that really meant The Gettysburgh Address. All of it. So when you get “4” in a text, that really means whole Gettysburgh Address. That’s compression (a stupid example).

In reality what happens is the MP3 system rips your sound apart and figures out the absolutely minimum amount of data you would need to put it back together at the other end. MP3 quality is not perfect. If MP3 thinks it can ignore some of the sound because you’re not likely to miss it, it gets ignored. That’s compression distortion. It gets worse as the MP3 quality value goes down.

That’s why MP3 is dangerous to use in production. It’s constantly dropping little pieces here and there, and sooner or later, you’re going to notice.


I just started recording and those were the specifications that came up. :blush:

So for recording a spoken address in a public hall, what specifications should I use? I see ‘Project Rate’ at the bottom of the screen. Should I set that at 16000 kH? As for the 32-bit float, where do I change that to 16-bit (or less)?

I just started recording and those were the specifications that came up.

Don’t need the red face. That’s perfectly valid.

The work window in Audacity has the shortcuts. The real settings are in Audacity > Edit > Preferences.

Attach 1
That’s where you set stereo.

Attach 2
That’s where you set Sample Rate and Bit Depth.

Those are my settings. Audacity will automatically convert to 44100, 16-bit on export unless you stop it.

My personal variation is 48000 instead of 44100 if I know I’m working for a video editor. Most of them don’t care, but it gives me a warm, fuzzy feeling.

Can’t have enough warm-fuzzy.

There is a place for super high sample rates and bit depths, but neither of us will be shooting Kanye West anytime soon, plus, that may take a more powerful computer than either of us own. Past that, those settings work really well.

Sub-sampling and lower bit depths are sometimes required for corporate telephone answering systems and other oddball, out of the ordinary applications.

Run away.

Screen Shot 2016-03-01 at 10.03.26.png
Screen Shot 2016-03-01 at 10.03.39.png

Love the bible, by the way.

Good work.


There’s that 32 bit confusion again…

32 bit is Audacity’s internal format. It’s higher than “physical” formats, to gain room for calculations. You can export in 32 bit, fi for mastering, to go to another DAW without converting. Besides that fringe case, you NEVER export 32 bit. A lot of simple players can’t play 32 bit anyways. And it takes twice the disk space.

When you “physically” record, your audio interface is either 16 or 24 bit. Those are the “physical” formats. As this is partly analog, you need to watch the noise floor. 16 bit gives you a (theoretical) noise floor of -96 dB. 24 bit gives you -128 dB.

When recording a podcast, use 16 bit. There is no need to go higher, as your source is 16 bit. There is no real “recording”, as the signal doesn’t leave the digital realm. It’s a copy from one bitstream to another. No loss and nothing to be gained by using higher bit depths.

When recording (multitrack) music to mix and manipulate, use 24 bit. That’ll give you room for mixing and effects.

When exporting for playback only, ALWAYS use 16 bit, either 44.1 or 48 KHz. The 48 KHz is for anything video. There’s no sense in “higher” formats.

The only reason you need to know all this, is because you need to avoid unnecessary conversions. Your computer will convert automagically between most, but not all formats. Windows Media Player, fi, has no problem with 32 bit, but has a lot of problems with 24 bit…

Thank you for all the help. Im getting the hang of it now. :slight_smile:

Just one more question. When exporting podcasts to MP3 (after recording at 44100 Hz) is there any advantage compressing the Hz rate to 22050 Hz?

A 22050 sample rate’s files are half the size when compared to 44100…

44100 has 22050 Hz, or 22 KHz as highest recorded tone. A bit less, really, as the filters need to be some distance from the top. Well beyond human hearing.

Half of the sampling frequency is the top of the frequency range.

22100 has 11 KHz as highest range. That leaves off a tiny bit of high tone. The recording will sound a little bit duller. Nothing to worry about as it is for use on a website and 99% of listeners won’t be using quality headphones or speakers.

There’s not really any benefit in reducing the sample rate when producing MP3s unless you are aiming for extremely small (and low quality) files. The MP3 algorithm is highly optimised to handle everything for you so you only need to be concerned with 3 things:

  1. Mono or stereo
  2. CBR or VBR (“Constant” or “Variable” bit-rate) see below.
  3. Size vs quality. Smaller size = worse sound quality.

For a podcast that is mostly speech, the biggest saving you can make is to do the show in mono and use the lowest bit-rate that still sounds OK.
The “CBR” (“Constant” bit-rate) mode is best for compatibility and generally recommended for podcasts. VBR tends to give sound quality that is a bit better, but some players display the playing time incorrectly or not at all. The “Presets” in the Audacity MP3 settings are VBR setting (the “Extreme” preset is excellent quality and well suited for high quality stereo music, but the files are relatively large for MP3s).

For speech recordings, “64 kbps CBR mono” is usually considered a good choice. If you need your files to be even smaller, you can go down to 32 kbps CBR mono, but the sound quality will suffer. I wouldn’t recommend going below 32 kbps because is sounds crap :wink:

Always do your production in high quality and save high quality backups (at least 16-bit 44100 Hz WAV). Encode as MP3 as the final step of the production after you have finished editing / processing and whatever else. Do not be tempted to re-compress an MP3 to make a smaller file. If you need a smaller file, go back to your original high quality version (or high quality backup) and make the smaller MP3 from that.

I did my last job at 16 kbs CBR. I’ll change it to 32 in this case. As for ‘mono’ the original recording was mono. However, the Export Options box only gave the choice of ‘stereo’ or ‘joint stereo’. I chose ‘joint’ but I wonder if a ‘mono’ button should be included there to avoid confusion?

If the Audacity project is entirely mono, then the export will be mono. That’s the same for all export formats. When exporting a mono file, the stereo / joint stereo option has no effect. For high quality stereo recordings, the “joint stereo” option generally gives better quality.

I’d expect 16 kbps CBR to be a bit low. I’ expect there to be a very noticeable swirly or bubbly metallic tone to the sound. It’s probably worth doing some test - get a nice clean recording and export it as 8, 16, 32, 64 and 128 kbps CBR, then load them all into Audacity and use the track Solo buttons to listen to one at a time and compare them. For a mono track, I’d expect the 128 kbps version to sound virtually identical to the original. The difference between the original and the 64 kbps version may be just a little more noticeable, but still pretty good. At 32 kbps you will probably notice that it “sounds like an mp3”. 16 kbps sounds pretty bad to me, and 8 kbps is dreadful. When doing this you will notice that for very low bit-rates, you will be prompted to select a lower sample rate - the LAME encoder will then convert the sample rate on the fly as it converts.