… for Codecs & Media

Tip #451: Audio Compression for Podcasts

Larry Jordan – LarryJordan.com

You can compress audio a lot, without damaging quality.

If you are compressing audio for podcasts, where it’s just a few people talking, you can make this a very small file by taking advantage of some key audio characteristics.

To set a baseline, an hour of 16-bit, 48k uncompressed stereo audio (WAV or AIF) is about 660 MB. (1 minute of stereo = 11 MB, 1 minute of mono = 5.5 MB).

If we are posting this to our own web site, streaming it live where bandwidth requirements make a difference, or posting it to service that charges for storage, we want to make our file as small as possible, without damaging quality. Here’s what you need to know.

Since people only have one mouth, if all they are doing is talking, not singing with a band, you don’t need stereo. Mono is fine.

This reduces file size by 50%.

NOTE: Mono sounds play evenly from both left and right speakers placing the sound of the audio in the middle between them.

According to the Nyquist Theorem, dividing sample rate by 2 determines maximum frequency response. Human speech maxes out below 10,000 Hz. This means that compressing at a 32K sample rate retains all the frequency characteristics of the human voice. (32 / 2 = 16K Hz, well above frequencies used for human speech.)

This reduces file size by another 33%.

Without doing any compression, our 660 MB one hour audio file is reduced to about 220 MB.

Finally, using your preferred compression software, set the compression data rate to 56 kbps. This creates about a 25 MB file for a one-hour show. (About 95% file size reduction from the original file.)

And for podcasts featuring all-talk, it will sound great.