Now that I'm finally able to use the application I developed in TeleFlow, I noticed some problems with the voice files.

I have high quality recorded WAV files (from 44Khz to 48Khz, Stereo) and used the Voice Manager to convert them to the VCE format.

The issues:

• One problem is that the final quality of the VCE files just sucks... Even VOX files are heard better on Dialogic hardware!
• Another problem is that some of the converted VCE files cannot be listened to at all (you hear some noise, similar to white noise, but with bumps).
• Finally, some of the source WAV files were not converted at all (the VCECOPY command gives error 0x3000c when ran manually).

The questions:

• I'd like to know what are the right or recommended PCM settings for the source WAV files, so the VCECOPY application can recognize them properly. I have a WAV editor/transformer so as to batch-convert them to a single quality.
• Also, I'd like to know what are the proper VCECOPY parameters to produce a higher quality VCE file, so as to use them in the NMSConvert.cmd batch file. Notice that I'm using a CG6060 board.
• Finally, what about the prompts? Do they get transformed as well (I just saw one of the WAV files under Prompts converted to VCE, but not the others, though I did not see any errors during transformation).

Thanks for your support.

The ideal recording settings are: 16-bit, 8 KHz, mono, PCM

You should record directly into that format for the best possible audio quality on an NMS board. Note that you do not need to convert to NMS VCE format. The board can play this PCM format, provided you have the "rvoice" DSP program loaded.

The NMSConvert.cmd program already has the only settings that are supported by TeleFlow. The 24 kbit/s compression was selected as our standard for .VCE support many years ago due to resource limitations. Those limitations are in the past, but the legacy remains. The real answer to your question is to use the raw PCM .WAV files instead, as per my advice above. This is the best possible quality that can be reproduced on that card.

The prompts do not get converted automatically by TeleFlow's Script Manager. The prompts are supposed to be PCM .WAV audio, and cannot be any other format than described above.

Ok, I converted the Scripts to the format you suggest and they actually sound better.

However, the Prompts are not converted to VCE by TeleFlow's Script Manager

I then tried to use them by adding the rvoice program like this:


... it took me awhile to optimize the resource pool sizes, but it's OK, the board starts just fine. However, the WAV file does not get played... This is an example of a successful run with VCE and the failure:


So, which 'rvoice' program should I load?

I believe this is the one you want: rvoice.play_lin

It works! Thanks a lot

Just curious... In the NMS documentation, where can I find more information about the specific DSP programs to be used for tweaking? I'm not touching anything now that it works, but I'm curious about how you conclude the the "16-bit, 8 KHz, mono, PCM" configuration is optimal.

Thanks again.

Well... "Optimal" is a subjective measurement. I will refer you to some NMS documentation on this subject, however you'll see in the newest NMS docs that your choice of codecs is expanded beyond what was available when our decision was made.

Firstly, the largest limitation is the effective audio quality that can be achieved on the PSTN. One source I had read, and I'm sorry I can't recall the name/location of it now (that research was done circa 1997 ), suggested that 8 KHz, 16-bit PCM was "best" for playing across an analog (POTS) line. The debates on this subject are almost "religious", as I've seen others claiming that 11 KHz, 8-bit is "best".

Secondly, TeleFlow supports 4 hardware "types": Ai-Logix, NMS, Dialogic and TF-Simulator (which is really a Windows media device). When selecting the optimal format to use, we wanted something that would be consistent across all 4 platforms. Unfortunately, the best match only supported 3 of the 4. At the time, Dialogic only supported PCM at 11 KHz, 8-bit, and NMS only supported PCM at 8 KHz, 16-bit. I haven't checked lately to see if Dialogic has added support for 8 KHz, 16-bit, but I did notice that NMS now supports 11 Khz, 8-bit.

Lastly, we wanted the default audio codec to be non-compressed, as most compression techniques lose a bit of quality. Raw 16-bit, 8 KHz seemed like a good choice at the time.

The best source of information on NMS's supported voice formats is "Voice Message Service Developers Reference Manual" (6422-24.pdf).

You can also get information from the individual board "Installation and Configuration" manuals, especially notice that the "rvoice" DSP program is relatively frugal when it comes to its MIPS consumption.