Before Buying Voiceguide

[briesmith]

Hello

 

We have installed a Dialogic 4PCI/Euro card in a Win XP PC and attached it to 4 Cat5 wall ports leading back to PBX extensions organised into a group of 4 pointing to one of our DDI numbers.

 

If we'd received any drivers when we bought the card (over £400) I suppose we might have been able to test our set-up but instead we spent all Friday downloading stuff (it's only 700mb or so) from the Intel website but in getting to grips with this project some questions have come to light.

 

Firstly, is the Dialogic stuff going out of production? We see different things on different websites, some of which indicate that it might be.

 

If it is, what are people (going to be) using instead? (I note your website still sort of recommends Dialogic).

 

If we buy VoiceGuide do we still need to spend US$1,000 on the Dialogic SDK?

 

What's a "line" so far as you are concerned? Is it an external PSTN line with its own diallable number or an internal extension driven by a PBX?

 

Given the set-up I've described above, what would we need to buy from you to allow 2 developers to work on building an application we could use to demonstrate IVR and our ability to build systems?

 

The demo would need to be able to take calls from outside our building - ie over the PSTN - capturing integer data from callers' keypads and also link to a database maintained by a remotely hosted website where any data gathered from the IVR script(s) would be collected.

 

The Dialogic 4PCI card doesn't support speech recognition. What card would/should we buy if we wanted to migrate our demo app to that capability? How much rewriting of the initial app would be required or would we have to start again?

 

I hope you can help with these questions.

[SupportTeam]

Firstly, is the Dialogic stuff going out of production? We see different things on different websites, some of which indicate that it might be.

No. Dialogic is the market leader in the area of telephony cards (especially analog cards) and they are continuously releasing new products. Some older products have recently went out of production but they were replaced with newer updated products...

What websites are you referring to?

If we buy VoiceGuide do we still need to spend US$1,000 on the Dialogic SDK?

No.

What's a "line" so far as you are concerned? Is it an external PSTN line with its own diallable number or an internal extension driven by a PBX?

Either type of line can be plugged into the Dialogic card (or modem)...

Given the set-up I've described above, what would we need to buy from you to allow 2 developers to work on building an application we could use to demonstrate IVR and our ability to build systems?

If you just have one D/4PCI system then just a 4 port license for that system is required...

The Dialogic 4PCI card doesn't support speech recognition. What card would/should we buy if we wanted to migrate our demo app to that capability? How much rewriting of the initial app would be required or would we have to start again?

VG right now does not support speech recognition, and it's a bit hard to say how many changes would be required to migrate apps to speech rec ones once that support is added, but it's safe to say at least you will bee required to add "Listen for Speech" blocks to the VG call flow.

[briesmith]

Thanks for the information. I'll let you know how we get on.

[briesmith]

Using a cheap microphone and a notebook PC to record my prompts. Got a lot of plosives and sibilants. Looked up CoolEdit to find its now Adobe Audition and costs £230. I thnbk it will definitely eliminate the Ps and Ss but before plunging in can you tell me whether it's worth it? Will the resulting quality improvements be noticeable/worthwhile given the quality of the average phone line and handset? Or would investing in a better microphone/sound card setup return better dividends? The .wav files are intended for a Dialogic D4PCI/Euro environment by the way.

[SupportTeam]

Professionally recorded sound prompts sound a lot better then what is usually achievable with 'sound card + microphone' recording.

 

To achieve good quality recordings you not only need a studio like environment - you much also have a pretty good voice... there's a reason why people who specialize in recordings are called 'Talents'.

 

For a list of a few 'Talents' please see:

http://www.voiceguide.com/servCallflowDesign.htm

[briesmith]

Problem with using talent is two-fold. Firstly, it costs a lot of money and second, you have to have a pretty iron cast, locked down script otherwise you've no sooner got your recordings back than you realise you've forgotten one!

 

For us, feeling our way by building our first IVR app as we are, I don't think using pro talent would be a good investment or approach.

 

There is also the very practical issue of not being able to speechify dynamically; you can't use talent to voice variables retrieved from a database, for example. So you end up with some very nice pre-recorded prompts followed by or interspersed with Stephen Hawking impressions (no offence intended).

 

I found a Scottish firm called rVoice which specialises in "sculpturing" (their word)TTS stuff. Their arguments for not using p[re-recorded prompts but relying on (their) better quality TTS technology are pretty convincing. Have you ever used them?

 

PS On a completely different tack; can you point me to an example script showing how a link to an ACCESS d/b is constructed? We want to get an ID from the keypad, look it up on an ACCESS d/b table and come back with either a .wav file to be voiced or a text variable to be "spoken" by a TTS engine. Sort of "Please enter your account no", "Thank you, Mr Smith, what can we do for you today?" Where the "Mr Smith" was found by looking up the d/b using the a/c no given by the caller. Any help you can give here would be much appreciated.

[SupportTeam]

I'd suggest trying the AT&T TTS engine - it's the best TTS around right now...

It costs $35 and can be purchased at:

https://www.regsoft.net/regsoft/vieworderpa...productid=56116

 

rVoice is OK, - have they gotten back to you with a price?

 

Regarding DB - please read the VG Help file section on the "DB Query" module, and take a look at the "Customer Enquiry" sample script.

[briesmith]

Thanks for replying.

 

I'm a bit confused about AT&T; on their website they say they don't supply their stuff for telephony applications?

 

I'm also a little confused about the difference between TTS engines as opposed to voices. I think all the engines do a good job, including good old "spawn of Satan" Microsoft's! What users hear depends on the voice and AT&T seem to want to charge big time for theirs.

 

Thanks for the pointer re d/b access.

 

PS The company I mentioned isn't rVoice - that's their product - it's Rhetorical (www.rhetorical.com).

 

PPS Do you know your email notification of replies isn't working?

[SupportTeam]

The $35 AT&T voice for SAPI 5.x engine (which is what VoiceGuide uses) is more then good enough for majority of applications of up to about 30 lines. You may want to read:

 

http://voiceguide.com/forums/index.php?showtopic=825

and

http://voiceguide.com/forums/index.php?showtopic=421

and

http://voiceguide.com/forums/index.php?showtopic=397

 

Products from Rhetorical and other similar are really aimed at installations with hundreds/thousands of lines - and their price reflects that. If you prefer to use rhetorical you can just buy their voice for SAPI 5.x and use it instead of the AT&T one if you want.

(Rhetorical is listed as one of alternative TTS choices listed on our downloads page...)

[briesmith]

Can't believe you've already responded! Have you guys ever thought about branching out into other areas; like trains say?

 

As I say I think it's the voice that counts; and the posts you kindly pointed me to seem to bear this out. Rhetorical's sculptured(?) voice does sound good but I hadn't picked up on the fact that you can buy it separately from the rest of their stuff. Thanks for the tip.

 

If I'm developing a script and I'm working at home where there's no Dialogic h/w installed will my script be transferable onto the Dialogic equipped m/c when I get back to the office?

[SupportTeam]

I'm working at home where there's no Dialogic h/w installed will my script be transferable onto the Dialogic equipped m/c when I get back to the office?

Yes. If you are testing at home with a Voice modem then you may need to record the sound files in a format preferred by Dialogic when you move the script - but that's about it...

 

Windows Voice Modem Wave driver can only support sound files in the following format PCM 8000Hz, 16bit, Mono.

 

Dialogic Wave driver prefers sound files in the following format PCM 11025Hz, 8bit, Mono.

[briesmith]

Rhetorical haven't been in touch so I can't tell you anything about (their) pricing - which I'm assuming might come as something of a shock?

 

Can't find an example script called "customer query"; is it part of one of the others;

Script 1 : Go directly to Voicemail Box 0001.

Script 2 : Auto Attendant

Script 3 : Credit Card Payments

Script 4 : Record message & deliver to specified number

Script 5 : Selecting Which Result Variable to use in later processing 

?

 

PS You know I don't think your email notification is working? Well, your "remember me" isn't working either. (I've got cookies enabled on my m/c and I've checked the email address on file for me and it's correct).

[SupportTeam]

The Customer Enquiry script can be found in VoiceGuide's \scripts\Customer Enquiry\ subdirectory.

 

Rhetorical is quite expensive...

 

Regarding using the forum:

1. We'll organize for the email notification of new posts on forum to be tested...

2. The 'remember me' seems to be working - we don’t need to log in every time...

[briesmith]

Sorry to be a nuisance but we don't have a /scripts folder; see the attached .gif. Where is it/do we get it from?

 

On the other issues; I have to log in each time and I don't get any email notifications. Because you reply so fast I'm checking back every 5 minutes or so!

[briesmith]

Found it - sorry being an a**e

[briesmith]

I'm trying (sensibly this time, I hope) to link two .wav files so that they play one after the other. How should |I enter this in the Selet Query - Sound File field?

 

The two files are C:\Program Files\VoiceGuide\Wavs\8bits\Scripts\LM_Info1.wav, Program Files\VoiceGuide\Wavs\8bits\Scripts\LM_Info1.wav

[briesmith]

I'm just cottoning onto some things regarding VoiceGuide.

 

To test a script I need a) at least a voice modem - there is no PC loudspeaker driving simulator - and an internal telephone extension - there is no keypad simulator, you have to "call" in to make the script work.

 

I'm using a pc with a built-in modem; allowing that I can disable it and attach an external (usb?) modem in its place, what one would you suggest (I'm in the UK)?

 

Will I need to prepare duplicates of all my .wav files because while

Windows Voice Modem Wave driver can only support sound files in the PCM 8000Hz, 16bit, Mono format, the Dialogic Wave driver prefers sound files in the PCM 11025Hz, 8bit, Mono format?

 

If I do is there a program that will prepare these duplicates, and with a batch option so I can do them all at once?

[SupportTeam]

We recommend external modems which use a serial connection - USBs don't work well... for a discussion on modems please see:

http://voiceguide.com/forums/index.php?showtopic=135

 

is there a program that will prepare these duplicates, and with a batch option so I can do them all at once?

Adobe Audition is pretty good for this...

[briesmith]

I'll take your most recent reply as both "YES" answers. Bloody nuisance about the dual format .wavs; especially as Adobe Audition is over £200 in the UK.

 

Can I repeat the query about linking .wav files; the recommended method - separate the path/names with a comma - doesn't work.

[SupportTeam]

Can I repeat the query about linking .wav files; the recommended method - separate the path/names with a comma - doesn't work.

Could you please post a copy of VoiceGuide's Debug Printout which captures the problem, this will allow us to see what is going wrong.

 

When running the script click on VoiceGuide's View menu and select 'Event Trace Log' option - any log information will then appear in this window. You must open the trace window before making the call.

[briesmith]

I will send the trace; just as soon as I can get a script to run!

 

I've got a modem set-up now and when I first switch it on it is set to auto answer. When I load VoiceGuide it initially says "waiting for a call" then I see the auto answer light on the modem go out. VB has turned it off; why?

 

When I do get VB to run it terminates a script by saying "waiting for a call" but then , some seconds later, it reports (the modem as) "unavailable" and I have to reset everything before I can make another call. All very frustrating and baffling!

 

I have to say your statements about "unreliable modems" can't really hold up. Millions of modems by reliable manufacturers are in use all day long throughout the world without any problems.

 

There would appear to be a little bit of laziness going on here - let's blame the other bloke - and things not being got to the bottom of.

 

We notice - frustratingly - that there is no modem reset button or command in your software (that we can find...); surely something like that would go a long way to sorting out this modem problem?

 

And a serial port diagnostic - where is it, what state is it in and who (which application) is it assigned to, would also help?

 

I don't think you can leave things where they are; viz "all modems are unrelaible; get a Dialogic board". I'd thought you'd want to get to the bottom of it?

 

I don't like to carp - your support has been fantastic - but it's been a very long day; each XP restart takes about 3 minutes...

[SupportTeam]

I've got a modem set-up now and when I first switch it on it is set to auto answer. When I load VoiceGuide it initially says "waiting for a call" then I see the auto answer light on the modem go out. VB has turned it off; why?

Actually the Window's modem driver (TAPI layer) would have turned it off, and then informed VoiceGuide that the modem is unavailable - that is why VG then displays "unavailable" in status window.

 

Sounds like there is something wrong with the modem's voice drivers... if you really want to find out what happened you may want to see what the modem driver has sent to the modem - you can do this using PortMon: http://www.sysinternals.com/ntw2k/freeware...e/portmon.shtml

 

The statements that modems are unreliable has more to do with the problems many modems have properly detecting DTMF/busy tones and sometimes falsely detecting DTMF/busy tones during playing & recording of sound files.

[briesmith]

Given up with the modem (Pro Phonic whoever they are); it starting crashing my XP m/c (not supposed to see the "Blue Screeen of Death" in an XP environment). Ordered another one, I'll let you know if that's any better.

 

The multiple file problem seems to revolve around playing them in script design mode; you can't. If you try you get an Invalid FileName error. Modules with multiple filenames seem to play OK when executing a script.

 

My query for today concerns block mode scripting. I want to script for:

- no answer was received - hangup;

- a star was entered - go back one step (module?);

- a valid response is 0 or 1, say, but some other value other then a star was entered - play a message reporting the error and request re-entry.

 

And I want this block on every module that is expecting directional (not values) input. I suppose I'm looking for an "else if" and an "else" construction.

 

How do I do this?

[SupportTeam]

The multiple file problem seems to revolve around playing them in script design mode; you can't. If you try you get an Invalid FileName error. Modules with multiple filenames seem to play OK when executing a script.

Once you have entered multiple sound files to be played you will not be able to play them as one file in the Script Designer - only when running the script.

 

With paths you will just need to specify in each module what to do. If you want the module to react to a particular key then you need to specify a path for it.

 

To go to previous module you can use the $RV_PreviousModule Result Variable.

[briesmith]

There are twelve buttons on a touch tone phone; does every module have to have 12 path statements; one for each possible key press? None of your example modules have every possible key press pathed. Is there no construction for "0" - do this; "1" - do that; otherwise - do this?

 

Without some sort of logical structure everything seems a bit longwinded and it makes it impossible to go back two or more modules.

 

EG A caller has entered three pieces of information as responses to separate modules but decides they want to redo their first response. How do you present an option that will allow them to go back and not just (endlessly) repeat the module they are in?

 

And why do you use the hash key as a delimiter in your examples? What purpose dose it serve? None of the modules I can find seem to test for it. Does it just get passed to some VB Script somewhere which can look for it and do something with it; ie is it transparent at the module level?

[SupportTeam]

Is there no construction for "0" - do this; "1" - do that; otherwise - do this?

There is no "otherwise" option. If a key is not specified in the paths then it is ignored.

Without some sort of logical structure everything seems a bit longwinded and it makes it impossible to go back two or more modules.

Usually the * key is used to go back through modules. Plz see the attached script.

A caller has entered three pieces of information as responses to separate modules but decides they want to redo their first response. How do you present an option that will allow them to go back and not just (endlessly) repeat the module they are in?

Just ask the caller if they would like to modify any of the information and just take them straight back to the module which they want to change their entry.

You can also design the script in such a way that the caller does not have to enter information in following modules again. See attached script.

And why do you use the hash key as a delimiter in your examples? What purpose dose it serve?

# is used to indicate end of entered data in Get Number modules where we are willing to accept any length number - caller uses # to indicate "end of entering the number"

Entering_Data_Demo.vgs

[briesmith]

Thanks for the example script; you're very kind.

[briesmith]

Well, we've got a modem and it works! It's (allegedly) an Intel chipset based Elite Series 56k Fax Modem and it will answer calls and play messages. So, so far, so good.

 

The only small fly in the ointment is that it won't work when connected to our Panasonic D1232 PBX via either the back of a telephone handset or to a wall port via a master adaptor. The wall ports are wired back to the switch using Cat 5 cabling.

 

The only way to get the modem to auto answer is to connect it directly to a BT wall port (ie teminating unit of an "outside" line).

 

Using another product I can force the modem to answer by clciking the "answer now" button. When I do this it picks up and answers the call fine when connected to the PBX but it won't answer of itself.

 

Any thoughts?

[SupportTeam]

The only way to get the modem to auto answer is to connect it directly to a BT wall port (ie teminating unit of an "outside" line).

Looks like the extensions of the Panasonic D1232 PBX system are not wired the same way as normal analog telephone lines - you should speak with your PBX technician to see what the differences are if they can change the wiring on this PBX extensions so that modem will be able to detect incoming ringing.

[briesmith]

Hello again

 

I've downloaded TextAloud and the Charles and Audrey voices and they work fine in TextAloud.

 

I've set-up some .txt files and pointed various PLAY modules to them so that the textbox contains the text file path and name where a .wav file would otherwise go. When I try and run the script, however, I get a voice (female) telling me that there is a file format error.

 

At the moment I'm set-up to use a voice modem and the play modules in the script (any script in fact) will quite happily play .wav files.

 

What am I doing wrong?

[briesmith]

Should have sent you the log!

 

164747.37 0 callstate OFFERING 66071 0 4

164747.37 0 Answer the call at 3/1/2004 16:47:47

164747.39 0 lineAnswer(66071) => 66054

164747.39 0 ring 1

164747.39 0 callstate ACCEPTED 66071 0 0

164750.70 0 tapi Reply (LineEvReply) ok 66054 0

164750.70 0 callstate CONNECTED 66071,0,0

164750.70 0 WorkingModeTAPI@Connected=

164750.70 0 WorkingModeScript@Connected=

164750.70 0 Inband detection not enabled

164750.70 0 StartLoadedVgs at 3/1/2004 16:47:50

164750.70 0 [Welcome] Playing

164750.72 0 tts generate start[Welcome to the leaguemanager.bizz results hotline. To enter your results press 1. To hear a short message describing how to use this service press 2. ]

164750.92 0 tts generate end

164750.92 0 [Welcome] Playing (C:\Program Files\VoiceGuide\data\tts0.wav)

164750.92 0 PlaySoundStart err=[0 play PlaySoundStart ResetWavOutAndContinuePlay=>7] file[C:\Program Files\VoiceGuide\data\tts0.wav]

164751.19 0 RunModule PLAY end

164751.19 0 tapie PlayEnd old line=0 0 569921

164755.84 0 Play End line[0] (id=569921)

164755.84 0 LsPlayMsg EV_PLAY_FINISHED

164801.20 0 LsPlayMsgFinished EV_TIMEOUT_REPLAYMSG

164801.20 0 tts generate reuse

164801.20 0 [Welcome] Playing (C:\Program Files\VoiceGuide\data\tts0.wav)

164801.20 0 PlaySoundStart err=[0 play PlaySoundStart ResetWavOutAndContinuePlay=>7] file[C:\Program Files\VoiceGuide\data\tts0.wav]

164801.47 0 tapie PlayEnd old line=0 0 580203

164806.09 0 Play End line[0] (id=580203)

164806.09 0 LsPlayMsg EV_PLAY_FINISHED

164811.45 0 LsPlayMsgFinished EV_TIMEOUT_REPLAYMSG

164811.45 0 tts generate reuse

164811.45 0 [Welcome] Playing (C:\Program Files\VoiceGuide\data\tts0.wav)

164811.45 0 PlaySoundStart err=[0 play PlaySoundStart ResetWavOutAndContinuePlay=>7] file[C:\Program Files\VoiceGuide\data\tts0.wav]

164811.73 0 tapie PlayEnd old line=0 0 590453

164816.34 0 Play End line[0] (id=590453)

164816.34 0 LsPlayMsg EV_PLAY_FINISHED

164820.13 0 callstate DISCONNECTED 66071,1,0

164820.14 0 LsPlayMsgFinished EV_REMOTEPARTY_DISCONNECT

164820.14 0 Hanging up call...

164820.14 0 RecSoundStop ok

164820.14 0 PlaySoundStop ok

164820.14 0 fnHangupCall end

164820.95 0 callstate IDLE 66071 0 0

164820.95 0 WorkingMode@Idle=

164820.97 0 tapi Reply (LineEvReply) ok 66037 0

164821.00 0 LsAwaitingCalls EV_TIMEOUT_TIMETOREINITLINE

164821.00 0 ReinitTelephony due to IDLE start

164821.00 0 tapic lineDeallocateCall(MainCall:66071) 0

164822.05 0 lineOpen(0) => 0

164822.05 0 Waiting for a call...

164822.05 0 lineOpen(0)LineHandle=66020

164835.89 0 sys cleanup Start

164835.89 0 sys cleanup End

165015.30 0 Loading VMB data begin

165015.30 0 Loading VMB data - about to delete old entries

165015.31 0 Loading VMB data - sorting/ordering

165015.31 0 Loading VMB data end

165015.31 0 Loaded new voicemail configuration file (XML)

 

Sorry.

[briesmith]

I've got the tts files to work now; I think the problem was caused by the text editor - NoteTabPro - I was using. I fiddled with the ascii/ansi output format setting and that seemed to make them work. I think before I did that there must have been illegal (formatting/eof) characters in the file or something?

 

What is happening is that I cannot input numbers from the telephone keypad while the tts files are being read which makes me think I've got a simplex (or half duplex) conversation going on. The modem allegedly supports full duplex but how you a) test for this or force it to full duplex I've no idea. Perhaps you can point me in the right direction?

 

Because of this keypad entries are "lost" and the whole script gets completely confused.

[SupportTeam]

Many modems do not detect keypresses well while playing a sound file - that is why Dialogic cards are used for professional solutions...

 

From the trace I don't think you are using v5.2 of VG - I'd recommend updating your install v5.2 as the tracing in that version can give us more info on what is happening with the system.

[briesmith]

I don't want to sound churlish but I wish you wouldn't keep going on about Dialogic crads. We've got a Dialogic card installed - and it works - but I want to develop on my machine not the server. Hence my use of a modem as a test bed which must make sense and be a very common approach?

 

It really isn't very helpful to keep on saying "but if you're using a modem what do you expect?" and the like. We are where we are and I might say that it would be more helpful if Dialogic and/or VoiceGuide recognised the needs for a development environment and provided better support for it.

 

I've been investigating the very poor sound quality I get - with a modem! - when using AT&T with the Audrey voice and I attach an email from technical support at Next Up (distributors of the voice stuff) which I thought you might find helpful and which really does suggets a response from yourselves.

 

Hello Brian

 

OK, that makes a big difference. I haven't tested with VoiceGuide, but most phone related applications have to put the audio out at 8khz. Audrey is a 16khz voice, so in the down sample to 8khz, it tends to sound a little crappy, especially S sounds, but others too.

 

Ken White

NextUp.com

The Power of Spoken Audio

http://www.NextUp.com

 

----- Original Message -----

From: Brian Smith

To: ken white - NextUp.com

Sent: Monday, March 01, 2004 3:19 PM

Subject: RE: AT& T Natural Voices

 

 

Hello Ken

 

It’s occurred to me that I may be completely misleading you.

 

I’m using AT&T’s engine with Audrey’s voice within an IVR application called VoiceGuide from an Australian company called Katalina. So I’m not actually using TextAloud to speak any of the TTS files.

 

I think this means that there won’t be any relevant *.dbg files to send you? If they are of any interest let me know and I’ll send them straightaway.

 

Brian

 

From: ken white - NextUp.com [mailto:kdwhite@nextup.com]

 

Can you check the demo at http://nextup.com/nvdemo.html using Audrey. Is what it does there drastically different than what it is doing on your machine?

 

Also, If you look in the TextAloud directory, you'll find a few debug files named TextAloudMP3.dbg, TextAloudMP31.dbg, TextAloudMP32.dbg, and TextAloudMP33.dbg. If you can email these files to me they should give me a better idea of what is going on.

 

From: Brian Smith

 

To: ken white - NextUp.com

 

Hi Ken

 

I don’t know how you’re getting on with Audrey but if you want to dial +44 74824153 you can hear her in action.

 

You will hear that her “s” sounds, particularly, are very poor and nearly all sound as “sh” sounds. They are much better when played through the sound carrd on the PC.

 

She also struggles with her intonation/phrasing putting peculiar emphasis on some syllables and she can’t say ”league” for some reason.

 

Any help you can come up with will be gratefully received.

 

Brian

 

I think you point out the difference in bitrates between the Windows drivers and those that Dialogic uses but I'm not sure I understand where we are regarding the frequency business. Perhaps you could help me out here?

[briesmith]

I've done a bit of searching around re half and full-duplex drivers and I've discovered that both are installed on my system.

 

Driver: Unimodem Half-Duplex Audio Device

Driver Type: Sound, video and game controllers

Driver Version: 5.1.2535.0

Driver Date: 1-26-1999

Manufacturer: Microsoft

ProviderName: Microsoft

Num files: 1

Files: C:\WINDOWS\Inf\modemcsa.inf

XMLIgnition: 3/2/2004 12:11:37

 

 

********************************************************************************

******

 

Driver: Unimodem Half-Duplex Audio Device

Driver Type: Sound, video and game controllers

Driver Version: 5.1.2535.0

Driver Date: 1-26-1999

Manufacturer: Microsoft

ProviderName: Microsoft

Num files: 1

Files: C:\WINDOWS\Inf\modemcsa.inf

XMLIgnition: 3/2/2004 12:11:37

 

 

The modem I have purchased purports to be a full-duplex modem. Do you know how the decision whether to install half or full-duplex modem support is made? Is it a function of PnP with the modem telling the OS what it is/needs?

 

Can I over ride the system choice and "force" the use of the full-duplex driver?

 

Currently System Info reports the following:

Name Intel V92 External (MD5660) Modem

Description Intel V92 External (MD5660) Modem

Device ID ROOT\MODEM\0001

Device Type External Modem

Attached To COM1

Answer Mode Not Available

PNP Device ID ROOT\MODEM\0001

Provider Name Intel Corporation

Modem INF Path oem15.inf

Modem INF Section Gen56kExt

Blind Off X4

Blind On X3

Compression Off %C0"H0

Compression On %C1"H3

Error Control Forced \N4-J1

Error Control Off \N0

Error Control On \N3-J1

Flow Control Hard \Q3

Flow Control Off \Q0

Flow Control Soft \Q1

DCB

Default <

Inactivity Timeout 0

Modulation Bell B1

Modulation CCITT B0

Prefix AT

Pulse P

Reset AT&F<cr>

Responses Key Name Intel V92 External (MD5660) Modem::Intel Corporation::Intel Corporation

Speaker Mode Dial M1

Speaker Mode Off M0

Speaker Mode On M2

Speaker Mode Setup M3

Speaker Volume High L3

Speaker Volume Low L1

Speaker Volume Med L2

String Format Not Available

Terminator <cr>

Tone T

 

WHich I imagine means something to someone?

[briesmith]

Trying to make progress here with developing a script but I'm being held back by two things.

 

Firstly, the script will sometimes accept and understand keyboard input and other times it won't. Module Get_PIN in the attached log file is the problem area.

 

Secondly, if I persuade the system to process input via that module as soon as I perss 1 to go on to the next module Play_Club_Name the script hangs up!

 

All very baffling.

 

This is the log

 

[support Team: pasted in log file transcript deleted and attached as .txt]

1.txt

Edited 03/02/2004 08:42 PM by SupportTeam

[briesmith]

Sorted out the second problem; it's caused by having two .txt files on the same play line within a module.

 

I created a second module, moved the second .txt file to it and the script works fine albeit with a slight pause between.

 

But only when I can persuade the script to accept the keyboard input. Sometimes it will, sometimes it won't. I can't work out why. Sometimes, if I press the keys in succession quickly it works; on other attempts it works when I add a hash sign to the end of the keypad input. On other tries both these methods fail.

 

What is going on? Any ideas?

 

How can I stop "your" American lady reading back the number strings entered? I don't need a read back as I look up the details entered and come back to the caller with a "welcome so and so" message if they've got it right. Otherwise I tell them I couldn't find them. All I want to say is "Thank you" and move on to the next module.

[SupportTeam]

I can see from your last posts that you are continuing to have problems with the DTMF tones being detected when playing sound file.

 

This is a well known problems affecting many modems.

 

As I said before the only option in these situations is to stop using that modem and try other hardware (... and yes, Dialogic will give you much better results).

[briesmith]

Thanks for the reply - and I'll be sure to attach log files as .txt in future, sorry about that - but as you'll see from my earlier posts I don't think it's a very satisfactory response.

 

The whole telephone system world wide depends on the reliable detection and interpretation of DTMF tones; without it we'd just get "wrong numbers" continuously.

 

Does the problem lie with the half-duplex driver I have? In the "old" days half-duplex (simplex) meant modems had to be "turned round" in order to reverse the data flow and latency was built-in to the systems to allow for the delay (typically 250ms) this entailed. Is this what is wrong here? Because the modem is operating in half-duplex, it has to be "turned round" from output, speaking mode, to listening, keypad accepting mode and this takes time. Time which isn't built in to the driver's buffering?

 

I really do need a definitive answer on the question of half versus full-duplex drivers.

 

I have commented in earlier posts on your insistence that we use Dialogic h/w, pointing out that it is bad practice to develop on a server; supposing we crash it or have to re-boot it? And, what would we do in a multiple developer environment; buy them all Dialogic cards? Most software development products recognise this "dilemma" by providing and supporting a development environment which, in my case, is to use "cheap" modems as a substitute.

 

Also, can you give some advice on the question in my last post concerning how to stop the American lady reading back the numbers entered (when I can get some in, that is)?

 

Finally, how come your American lady's speech is so clear while Audrey's - a much more expensive lady I'm sure - is so poor?

[SupportTeam]

The whole telephone system world wide depends on the reliable detection and interpretation of DTMF tones; without it we'd just get "wrong numbers" continuously.

... That is why hardware which can detect DTMF tones reliably must be used... Looks your modem does not detect DTMF tones reliably... that's pretty much it.

 

Does the problem lie with the half-duplex driver I have?

All modem’s wave drivers are 'half-duplex' - meaning they can only play or record at a time - not both at the same time. But that has nothing to do with DTMF detection - which happens (not very reliably on some voice modems) regardless of whether the modem is playing or recording or neither.

 

how to stop the American lady reading back the numbers entered

VoiceGuide will stop playing of sound files if a DTMF tone which has a path assigned for it is detected on the line...

 

Finally, how come your American lady's speech is so clear while Audrey's - a much more expensive lady I'm sure - is so poor?

The prompts shipped with VoiceGuide are pre-recorded in a studio. The prompts TTS system generated prompts will never be as good...

[briesmith]

If all modem drivers are half-duplex why - as the print-out I sent in an earlier post shows - do I have both a half and a full-duplex modem driver on my system?

 

How can I make the system use the full-duplex one instead of the half-duplex driver?

 

Why don't/can't you answer the question I raise - which is a fair one - about development environments? Surely you cannot expect every developer - there could be 10s or more on a big project - to have their own Dialogic card?

 

VoiceGuide's stance on modems and their intrinsic unreliability is simply not supportable.

 

You misunderstand my question about the American lady; I'm concerned with the logic necessary within a module to prevent her being invoked at all. Having had a number string entered I just want to go onto the next module with no read back.

 

Thanks for putting me straight about the voice quality. One thing intrigues me though. How come there is no delay between her saying "you have entered" followed by the numbers input being read out? I can't get consecutive modules to execute that seamlessly.

 

And, finally, if these are voices recorded in a studio why didn't you get Kylie to do them? ()

ModemDrivers.txt

[SupportTeam]

I suspect the full duplex driver would have been installed for the "Simultaneous Voice and Data" which is supported by some modems. This is very different to playing & recording sound files on the telephone line.

 

Voice Modems cannot play and record at the same time. This is just the way they are designed.

 

 

The reality is also that not all voice modems are reliable - as you are finding out - and any complaints about that should really be directed to the company which made the unreliable modem. Other posts on this forum (or even just the previous posts in this thread), in the Knowledge Base and Help file make this point many times. Since the modem you have is defective then the usual course of action is to return the defective item and buy a different model... read: http://voiceguide.com/forums/index.php?showtopic=135

 

Most IVR projects do not require a large number of developers for the actual IVR programming - actually just 1 developer is most common, with a great majority of scripts taking less then a week to fully design and deploy.

And yes, in professional environments the IVR developer usually do have their own Dialogic card... very few people in professional environments use voice modems.

 

If you do not want the number to be read back after it is entered then you should not enable the "Confirm Entered Number" option in Get Numbers module.

[briesmith]

I've surrendered and set myself up to use the Dialogic card so no more on the modem issue from me, I promise. I did manage to "force" Windows XP to install the full-duplex driver but I couldn't discern any difference and key presses were still missed so I've accepted defeat.

 

Audrey through the Dialogic sounds better than through the modem but because she's downsampled from 16Khz to the Dialogic 11.25Khz format "on the fly" she doesn't sound her natural(!) self and you get the occasional splodge of noise as well.

 

Do you know if there are any Dialogic preferred format voices using UK English?

[SupportTeam]

Do you know if there are any Dialogic preferred format voices using UK English?

There is no set of pre-recorded VoiceGuide system prompts in UK English. If you must use UK English instead of the US prompts which ship with VG then you will need to re-record the prompts which will be used.

[briesmith]

You misunderstand me, my question concerned voices - like Mike, Crystal, Charles, Audrey etc - not pre-recorded prompts.

 

I was trying to locate - if any exist - voices that actually correspond to the Dialogic preferred format (11.025Khz etc) as, I assume that when Dialogic chose it they did so for a reason and not just out of perversity? The only company I can find, so far, is Rhetorical who,as we all know, are a bit expensive.

 

My other question - about using Kylie as opposed to some anoymous middle-aged American lady- was an attempt (failed) at humour, sorry.

 

Seriously, I would have thought that Cary Grant or Elizabeth Taylor voices would be the best approach. They are "American" enough to be understood in that country while being sufficiently neutral as to not raise hackles elsewhere in the English speaking world.

 

I'm surprised if you can get away with using an American voice for IVR applications in Australia. People hearing American sounding prompts in the UK think they've got a wrong number and hang up in case they're being charged for a transatlantic call!

 

PS I see you've finally fixed your cookie and posting notification problems; thank you, it was getting really irritating.

[SupportTeam]

After testing all available TTS voices available we came to a conclusion that from the affordable ones AT&T sounds best.

 

Other people who tested a number of TTSs as well came to the same conclusion.

[briesmith]

I think AT&T are the only alternative if you want a UK sounding voice and can't afford Rhetorical.

 

The big problem - and it really is a showstopper - is that Charles and Audrey sound so crap over the 'phone using Dialogic cards due, I imagine, to the downsampling from the original 16Khz to Dialogic's 11.250Khz. We've moved to a Pentium 4 3.2Ghz m/c with 1Gb of memory running Win XP to try and make sure the fall-off in quality wasn't due to processor overload.

 

Listening to the Charles and Audrey demos on the NextUp website and then giving our script a go you'd never think you were listening to the same stuff.

 

Once our script is bedded down I think the only realistic option - if we want anybody to listen to it, that is - will be to re-work it substituting pre-recorded voice files for all the TTS stuff in it at present.

 

That's a lot of work and leaves us with what to do when reading back unforseeable (dynamic) variables. Lapsing into your American lady's voice as a default is really not an option.

 

Does nobody make a Dialogic equivalent that supports 16Khz?

[SupportTeam]

... most people who listen to AT&T over Dialogic judge it to be quite good and easily understood by caller.

 

Other readers of this thread who want to hear what AT&T sounds like can download this file: http://www.voiceguide.com/support/tts7.wav which was generated using AT&T and play it on their system - the sound quality achieved will be the same as when the TTS is generating the sound files live on your system.

 

I do not know of any cards which would play at 16kHz on a telephone line.

 

[21 Sept 2006 : Updated link to WAV file]

Edited 09/21/2006 02:27 AM by SupportTeam

[briesmith]

I am surprised that you think the audio quality is acceptable and you don't comment on just how far removed it is from the playback quality you get using the demo facilities on the AT&T website (I gave the URL in a previous post).

 

I suppose it's all down to personal taste and expectations but there does seem to be a long way to go before TTS on telephone systems matches anything achieved using ordinary PC sound set-ups.

 

As a first step it would make more sense if the VG default voice was the declared application voice so that when VG spoke variables you didn't get such a clash between, in our case, our declared application voice of AT&T's "Audrey" and your default American lady.

[SupportTeam]

Phone lines will never sound as good as sound coming from speaker on your desk.

 

To start off the phone lines' frequency response is limited to about 3.5kHz, and the speakers used in most handsets are also nowhere near as good as most PC speakers.

 

Any sound files, whether or not they are TTS generated will not sound as good on phone lines as they do through PC speakers.

 

If your are using the TTS engine then you can just use the TTS engine to speak everything. There is no need to use "Say Number" modules, the