Yeah, this one comes up fairly frequently… There’s some new information to cover here, however, so let me start from the top.
One of the things that confuses folks is the size of the file created for a voice mail vs. the sampling rate used to record it. The sample rate is how many data points are recorded as you speak… the higher the sample rate the better the sound quality (in general) and the larger the resulting file. The amount of storage the resulting WAV file uses depends on what kind of compression the codec uses (if any). The higher the compression, the smaller the file and the smaller the “pipe” needed to stream the message across the network needs to be… however you generally sacrifice some sound quality and, more importantly, some processing cycles are spent compressing at record time and then uncompressing at playback time in most cases. The analog cards used in Unity (NMS, Dialogic) have the ADPCM (mentioned below) compression/uncompression functionality built in at the hardware level which means our CPU doesn’t need to waste any cycles dealing with it, which is nice.
OK… now the real question is what sampling rates/compression capabilities can Unity handle. Unity 2.4.5 supports the following formats:
MuLaw: 8K/sec sample rate, 8K/sec storage. No compression, sounds the best and is the default codec out of the box.
ADPCM 6K/sec sample rate, 3K/sec storage. Hardware based compression if you’re using analog cards. If you’re using the AvCiscoTSP with CM, the compression/decompression happens at the CPU level… proceed with caution.
ADPCM 8K/sec sample rate, 4K/sec storage. Again, it’s hardware based compression for analog cards.
Alaw: 8K/sec sample rate, 8K/sec storage. No compression. Similar to MuLaw, used mostly in Europe.
G729a: 4K/sec sample rate (about), 1K/sec storage (roughly). Amazing compression and very light on the CPU. The sound quality is not as good as MuLaw by any means but most folks find it very acceptable.
So… doing some quick math you come up with:
MuLaw and ALaw – a 1 minute message is roughly 480K
ADPCM 8K – a 1 minute message is roughly 240K
ADPCM 6K – a 1 minute message is roughly 180K
G729a – a 1 minute message is about 60K.
There’s a small amount of email header space involved for each message but the numbers above generally are accurate and you should be able to reproduce them yourself.
As mentioned several times out here, if you change the default codec on the system from MuLaw to any of the other formats up here, you need to be aware of the consequences. It’s a system wide change and affects ALL recordings, including greetings and voice names, not just messages. If messages are being forwarded to folks who do not have VMO installed on their desktop, they will not be able to play these messages unless they download the codec and manually install it. In the case of ADPCM you can get that off our web site. In the case of G729a, that’s not the case… it’s not a free, publicly available codec (i.e. some company gets a few bucks for every install and we have to shell out to them for every CD we ship starting with 2.4.5). The lawyers are still thumb-wrestling over that one, I’m sure something will get worked out.
Either way, you have to edit the registry to change the default codec and restart Unity. Open your favorite registry editor and head to "HKLM|Software\Active Voice\MIU\1.0\Initialization". You'll see a key there called "Default WAV Format". It'll be set to "3" for MuLaw. For 3K ADPCM set it to 1, for 4K ADPCM set it to 2, I think G729a is 5, etc… you’ll see the list of codecs in the WAV formats folder there and you can find the one you want. As usual, edit the registry at your own peril… please know what you’re doing and proceed with caution.
Jeff Lindborg
Unity Product Architect
Active Voice
jlindborg@activevoice.com
http://members.home.net/jlindborg
