PhoneBoy Explains: Internet Audio

Here's a brief summary of the various audio formats used on the Internet. Where possible, I've included a voice and a music sample. They are all the same so you can hear them side-by-side and make your own conclusions. With stuff that I can not serve off of my web server, I've included links to sites that contain appropriate content. If I missed something or you want to send feedback, email me at phoneboy@phoneboy.com.

Explanation

High-quality audio on computers takes up a lot of space. For example a 22 second sample from one of my favorite CDs would take up almost 3.9 megabytes of disk space on my computer. This is sampled in stereo with 16-bit resolution at 44,100 times per second (also known as 44Khz)-- the same as is stored on the CD itself. This would take anywhere from 20 to 40 minutes to download from your modem. At a noticable loss in quality, I can lower the number of samples per second (8000, or 8Khz) and turn it into a mono file. Even after I've done all that, the file is still 345k, which takes a 2 to 4 minutes to download.

Aside from the sheer size of audio files, there's the problem of allowing people to hear your audio "as it happens" or even "while it downloads" to prevent having to wait for the entire file to download. Several companies have different products to solve all of these problems. The products are implemented in a variety of ways and use a variety of different CODECs (Compressor/Decompressors) which make the audio files smaller at the expense of some audio quality.

The most popular and well-known of these products is called RealAudio, which you may or may not be using to listen to this report. RealAudio is a general purpose audio format for the Internet. Back in 1994 when it first came out, it allowed for pre-recorded audio to be "streamed" to a person's computer over a 14.4k modem connection. Later they added the ability to serve up "live audio" (like a traditional radio broadcast) over modem connections. More recent versions fully utilize 28.8k and ISDN modem connections. The greater bandwidth allows for higher quality audio to be transmitted to the computer. Other general purpose audio formats include TrueSpeech, which is optimized to work at 14.4k modem speeds, Internet Wave, which can be optimized for anything up to 28.8k modem speeds, and Streamworks, which can be used on anything from 14.4k modems to ISDN lines.

Voxware decided to take a slightly different tact in it's first audio product. They decided to optimize their CODEC for speech, which makes for smaller audio files, but on anything other than speech, it sounds horrible. And, though the speech is listenable, it sounds like you're talking thru a tin can. But it is very low bandwidth -- it generates files that can be streamed over a 2400 baud modem! Voxware is working on a more generic CODEC that generates files that are still smaller than your average RealAudio or TrueSpeech files.

Which one is better? I've digitized some audio, both voice and music, listed the pros and cons of each format, and set them up on this web page. You can determine for yourself which one is better.

RealAudio

TrueSpeech

ToolVox for the Web

Internet Wave

Streamworks

Last Update: 18 July 1997
Return to PhoneBoy's Internet Guide