MiContact IVR Prompt Quality Question

[Devolution]

Hi All,

I am aware of the standard settings for WAV files in the IVR prompts i.e. 8000hz, mono, u-law but does anyone have any tips to stop it from sounding like a phone call from the 80's? If I can have a crystal clear call between internal handsets using G711, why can't I have the same for an external call coming into the IVR using G711? Why is the IVR prompt the weak link or am I missing something?

Thanks.

 

[Valamagules]

I suppose it really depends on how you get the greeting recorded.
I have seen a huge difference between greetings that are professionally recorded compared to the ones that arent.

Also may want to check QoS on the RTP from the MiCC to the voice network to make sure that it's correctly tagged for Voice (EF / DSCP 46).
I have seen in the past that MiCC (being a Windows Server) sends traffic as untagged hence no QoS.
To achieve this, you will need to implement this on the GP Policy of the server to tag Voice traffic correctly (similar to how you do it on a MiCollab Softphone).




Clever men learns what Wise men shares!

 

[Devolution]

Thanks Valamagules. I think you might be onto something with the QoS settings. I doubt this has been set up from the MiCC server. I will investigate this on our in-house solution.

In regards to the recordings, I tested this by recording some prompts in my home studio setup with a Shure SM58 microphone into Cubase. I applied some vocal compression to boost lows and restrict highs, then used some EQ to eliminate any low or high-frequency bands not used. The idea is that when it comes to compressing it down to Mitel values, it only has to compress what is required.
After all of that, I didn't really hear a noticeable difference. Only slightly fewer "S" sounds on the high frequencies. It would be good to know what professional companies do to make it sound better. Possibly the equipment but more likely decent mastering of the final audio file.

 

[jpruder]

This makes me go back to my school days. Bell labs determined that telligible voice only requires bandwidth from 300-3600 (some might have a different upper boundary). This is why the mu law sampling rate is 8000, Nyquist twice the highest frequency. The brain is awesome and adds harmonics even if not within the bandwidth.