Asterisk IVR has very poor voice quality

For comparison purposes, I set up two very similar IVR systems. One is
a traditional IVR based on a Dialogic voice board while the other is a
software-only IVR based on Asterisk.

The quality of the voice in the Asterisk IVR is awful.

If I add some hardware, will the quality improve??


that’s like saying “this new car is slower than my other car. why ?”

we’ll need to see some specs ! asterisk server, network, other services and phones attached please.

and i also want to know if there’s some method to improve the music quality , a 128 kbps mp3 sounds terribie

[quote=“baconbuttie”]that’s like saying “this new car is slower than my other car. why ?”

we’ll need to see some specs ! asterisk server, network, other services and phones attached please.[/quote]

The server is a Dell which has plenty of resources (multi CPUs, gigabytes of RAM), and it is not being used for anything else than Asterisk at the time. The network is a quiescent Cisco gigabit switch, at night when it has no users. I must stress that this is NOT a PBX, it is simply running a test message that says something like “Press number 1 if this and press number 2 if that”. It has no telephones attached. I just call from my home phone to the office to hear the quality of the 2 IVR systems. The call is received through an ISDN PRI, and a Cisco AS-5300 sends the call, converted to VoIP to the Asterisk server, and through POTS telephony to the Dialogic IVR server.

I have been deploying IVR and calling card systems based on Dialogic and SCO for about a decade. These systems handle thousands of calls per day. All I want is to get rid of SCO and use Linux, but so far I am not too impressed by the voice quality (using default codec) of Asterisk as an IVR.

In the Usenet VoIP newsgroup somebody said that if I attach a zaptel board the quality of the voice should improve. Can somebody confirm this?


The quality of audio from the pre-recorded prompts is based on so many factors it’s hard to determine if a single action (say, installing a Digium card) would address your concerns.


The quality of the microphone/studio used to record the prompts.

The sample rate of the prompt recordings.

The compression rate of the file.

The ability of the DSP or processor to keep up with the playback.

The quality of the phone or speaker used to hear and compare the two.

Just saying, “buy a Digium card” is, I think, a little simplistic. The first thing to do is be certain all things are equal.

You should know the sample rate and compression of both files you are playing, remembering that a severe compression rate and a low sample rate would degrade sound on either system.

You should remember that if you test a Digium card with a single port in service the processor may keep up nicely, but may stutter and fail with all ports in service.

Engineering systems like this is always a balancing act. Yes, it’s nice to have the best sounding prompts in the business, and I think it’s good that you take the time to compare things like that.

However, if you have to buy the most expensive boards and have huge storage for prompt files it increases your cost to deploy a system.