MP3 Speech sounds bubbly at 96kpbs

[asloane]

I need to convert lectures from their original 16-bit WAV and 320 kbps MP3 format down to 96kbps MP3's for audio streaming.

When I do this in Audacity 1.3.9 beta (and previous versions) I am hearing some minor distortions, which were not there before, which sound like the lecture is happening underwater. I had assumed that 96kpbs was adequate for speech but in these cases it is very far away from perfect. It's worse at lower kpbs rates, so I am afraid to go any lower even though a lower rate would be better for listeners with slow connections.

Is there anyway to avoid this situation and produce better quality MP3's?

[kozikowski]

It's true. 32 is the minimum for clear Mono MP3 and 64 for Stereo..... Unless you started with a compressed master. Then the errors add up.

Those numbers only work if you start with a clear, stereo, high quality, uncompressed master sound file.

MP3 is a delivery format. WAV is a production and editing format. The two formats don't cross well.

Koz

[steve]

I need to convert lectures from their original 16-bit WAV and 320 kbps MP3 format down to 96kbps MP3's for audio streaming.

You will get much a better sounding MP3 if the recording is mono rather than stereo. If you have a stereo track, convert it to mono before exporting.

If you use Audacity 1.3.x you also have the option to export as Variable Bit Rate (VBR) MP3, which will generally give slightly better sound quality than Constant Bit Rate (CBR). The downside to VBR is that some MP3 players will incorrectly report the length/playing time (though almost all MP3 players these days will play VBR OK).

For speech you may find that it sounds less bubbly if you filter the recording before encoding - (very easy to do in Audacity 1.3.x). For intelligibility, speech is almost all in the frequency range of 300Hz to 3kHz. If you use a 12dB/octave High-pass filter set to about 150Hz, and then use a 12dB/octave Low-pass filter set to about 8kHz (10000Hz) you will probably find that it sounds pretty good right down to 64kbps CBR (mono).

[asloane]

Those numbers only work if you start with a clear, stereo, high quality, uncompressed master sound file.
Koz

I can certainly master in WAV although it requires more storage space. When I did convert from the WAV original to MP3 I was still getting the bubbly sounds at 64kbps

[asloane]

You will get much a better sounding MP3 if the recording is mono rather than stereo. If you have a stereo track, convert it to mono before exporting.

Thanks. I will record in mono.

I don't know whether to use VBR because of my Shoutcast limit and listener slow connection speeds.

I am currently streaming in Shoutcast for which there is a maximum rate, although I have received a higher CBR than the system spec e.g. my limit is 64 but i have received 128kpbs. If I convert to VBR then will this cause problems with Shoutcast? The other consideration is that some listeners have slow connections and experience stuttering in their listening at rates above 80kpbs, so I assume that VBR would cause problems unless I can set the upper limit at 80-96kbps?

[steve]

CBR is the best for compatibility, so I would probably use 64kbps CBR (mono) for speech.