Document Reference

Technical Support
Intervoice wants its customers to be satisfied with its products. Your service contract and/or support guidelines specify which regional support office to contact. If you need assistance, contact Intervoice technical support at the numbers listed below.

Western Hemisphere

For 24-hour support, please telephone or fax the following regional support office:
(outside USA) (within USA)
Tel: +955-4688 or +454-8130 Tel: (800) 955-4688 or (972) 454-8130

Eastern Hemisphere

During normal office hours, please telephone or fax one of the following regional support offices:

England

(outside UK) (within UK)

Germany

Tel: +Fax: +Tel: 1234 Fax: 1007
(outside Germany) Tel: +Fax: +(within Germany) Tel: 440 Fax: 4444

Singapore

(outside Singapore) Tel: +4300 (within Singapore) Tel: 0395 4300

United Arab Emirates

Fax: +4277 Fax: Fax: +5580 Fax: 5580
(outside UAE) (within UAE)

Tel: +5480 Tel: 5480

If out of hours support is included in your contract, please telephone: (outside UK) Tel: +(within UK) Tel: 6387 Out of hours calls are directed to a support desk, where you can record your problem details. You will be contacted as soon as possible.

In this chapter.

Overview
The Sound Forge and the Intervoice System Audio File Guide provides information on creating audio files using Sound Forge. Topics include studio set up, file formats and file naming conventions, post-processing using the Batch Converter, and troubleshooting. This document also provides detailed instructions on using the CRVoice2 utility to rename audio files to the Intervoice runtime format. Sound Forge can be used for recording and editing, but it is required for voice file postprocessing, noise floor testing, and complex editing. The audience for this guide includes voice talents and speech developers interested in an efficient method of postprocessing voice files, testing the noise floor, and editing complex voice recordings. This chapter provides the overview for this document. It contains a purpose statement for the entire document. It also contains a concept diagram and a list of required hardware and software. For more information on preparing voice files, refer to the Sound Forge documents and Intervoices InterSoft runtime documents. For more information on Sound Forge, visit theWeb site: http://mediasoftware.sonypictures.com/support/, or visit the Intervoice Web site at www.intervoice.com/support/techinfo/voicerecording.
Summary of Changes, 1-2 Voice Recording Concept Diagram, 1-3 Recommended Hardware, 1-4 Recommended Software, 1-5

Summary of Changes

The major revisions to this document are summarized below:

Changes

Updated the Web site URL for downloading Sound Forge and Batch Converter. Reorganized chapters into a more functional order. Replaced Omvia Media Server term with Intervoice runtime system (or more specifically with InterSoft IQTalk runtime system or VoiceXML Browser system).

Additions

Added information about Sound Forge version 7. Added information about recommended audio file formats for VoiceXML Browser systems.
Voice Recording Concept Diagram
Figure 1-1 illustrates the process of creating voice recordings.

Computer

Voice Recording Software

Recording

Preamplifier

Speakers Microphone

Sound Card

Master.wav files

Processing
Sound Forge Batch Converter
Intervoice format.ivc files for InterSoft IQTalk
Runtime.wav files for VoiceXML

Renaming

CRVoice2 Utility

runtime files

NA for VoiceXML

Intervoice System

Figure 1-1. Voice Recording Concept Diagram

Recommended Hardware

The following hardware is recommended: Intel Pentium or compatible computer Sound card Preamplifier Amplified speakers with cables High-quality, unidirectional, low-impedance microphone XLR Cable (connects the microphone to the preamplifier) Y-adapter audio cable (connects the preamplifier to the sound card)

Specific Hardware

Intervoice recommends the following hardware for the best quality when recording: M-Audio Audiophile 2496 sound card Midiman Audio Buddy preamplifier Sennheiser E845 supercardioid vocal microphone
Note In noisy environments, a head-mounted,
noise-canceling electret microphone, such as the Shure WH20XLR, is recommended. Microphone base Horizon 10-ft. XLR microphone cable (connects the microphone to the preamplifier) 6-ft shielded audio cable (1/4-inch to RCA jack) (Radio Shack 42-2373A) 6-ft Y-adapter audio cable (1/8-inch to RCA (phono) or 3.5 mm stereo plug to two RCA plugs (connects the preamplifier to the sound card) 3.5 mm stereo jack-to-jack shield adapter Labtec speakers (AC powered)

Recommended Software

The following software is recommended: Sound Forge 6: Microsoft Windows 98, ME, or 2000 Intervoice CRVoice2 utility Sonic Foundry Batch Converter 5.0
Sound Forge 7: Microsoft Windows 2000 or XP Intervoice CRVoice2 utility Sonic Foundry Batch Converter 5.0
This chapter describes the elements necessary to properly set up the recording studio. It explains how to install the voice recording hardware and configure volume controls. It also discusses how to set recording attributes and the microphone level, and how to perform the noise floor test.

Studio Setup Checklist, 2-2 Installing the Hardware, 2-3 Installing the Sonic Foundry Batch Converter, 2-6 Installing CRVoice2, 2-7 Installing Sound Forge, 2-7 Making Quality Recordings, 2-8 Performing the Noise Floor Test, 2-12

Studio Setup Checklist

Use the checklist in Table 2-1 to ensure that you set up the studio properly.
Table 2-1. Studio Setup Checklist Studio Setup Checklist

Set up hardware.

Install the sound card, following the manufacturers instructions. Connect the preamplifier. Verify that the preamplifier is attached to a line-level input from the sound card. This may be labeled line or aux. Connect the microphone.
Note Using a penny, remove the gold plated screw in the microphone mount.
Then attach the microphone mount to the top of the microphone base. Connect the speakers. Turn on the computer. Refer to the OEM documentation for installation instructions. Turn on the preamplifier.
Install software. Sonic Foundry Batch Converter 5.0 CRVoice2 utility Sound Forge 6.0 or 7.0
Set up audio. On some sound cards, you must select correct audio input using the sound card control application. The recommended sound card, M-Audio Audiophile 2496, does not use an audio control application on the PC. In this configuration, the audio volume is controlled through the preamplifier. Set up audio levels. The input level should be set as loud as possible without clipping the.wav file. Perform noise floor tests.

Installing the Hardware

The following hardware is required for voice recording:
Sennheiser E845 Microphone

XLR Microphone Cable

Midiman Audio Buddy Preamplifier
Audio Cable 1/4-inch to RCA Jack Power Cord
Jack-to-Jack Shield Adapter

Y-adapter Audio Cable

Labtec Speakers AC Powered with Cables
M-Audio Audiophile 2496 Sound Card
Figure 2-1. Voice Recording Hardware
Connecting the Sound Card to the Speakers

www.intervoice.com/support/techinfo/voicerecording.
Save the file to the directory where you will be converting your.wav master files to.ivc files for InterSoft IQTalk systems.
Refer to Using the CRVoice2 Utility on page 6-4 for more information.

Installing Sound Forge

To install Sound Forge:
Ensure that the Batch Converter, sample Intervoice script, and CRVoice2 utility have already been installed. Insert the Sound Forge CD-ROM in the CD-ROM drive. The installation program should autostart. Follow the installation instruction screens. Go to the Web site and download the latest update to Sound Forge:
http://mediasoftware.sonypictures.com/download/step2.asp?DID=4 65. (This update fixes an.ivc file issue.)
Run the downloaded file (such as soundforge70b.exe) and follow the onscreen installation instructions. Refer to the Making Quality Recordings on page 2-8 and the other topics in this document for information about using Sound Forge.
Making Quality Recordings
To ensure that recordings are of the highest quality, set the microphone levels correctly, place the microphone properly, and avoid clipping. As a general rule, microphone levels should be as loud as possible without causing any clipping. You can adjust final volume levels in message files after they have all been recorded. Use the volume or the dynamic graphic preset. The best signals in the noise ratio are achieved when a loud, unclipped recording is made.
Setting Microphone Levels
To set microphones levels correctly, follow these steps:
Set the recording attributes using File > New as 11,025, 16 bit, Mono. (Refer to Setting Recording Attributes on page 4-5). In the main Sound Forge window, select the Record button the left of the button bar. The Record window appears. on
Set the microphone volume to the highest level at which you can speak at a normal volume without clipping. This produces the best signal-to-noise ratio.
In Sound Forge 6, check the Monitor checkbox beneath the Reset button to turn on the record meters.
Speak into the microphone in your loudest normal voice. Adjust the microphone volume on the preamplifier so that clipping does not occur even when you speak at your loudest volume. (Clipping occurs when the level meters are pushed into the red zone and the word Clip appears.) If clipping occurs, click the Reset button. Remember, if clipping occurs during the recording process, the recording is ruined. On the preamplifier, lower the microphone level. Repeat step 4 through step 6 until clipping no longer occurs. The microphone level is then set correctly.

DC Offset removal

Removes any DC offset before messages are compressed. Most sound cards have some DC offset that can affect the compression process.
Front and Back silence trimming with 20-ms fades
Removes excess silence on the front and back of messages to tighten the message delivery. Generally the messages are trimmed to 20 milliseconds or less of silence on both ends. Noise at the beginning or end of messages (such as mouse or keyboard clicks) causes the auto-trimming process to trim too early or too late, leaving extra silence in the message. These messages must be hand-trimmed.
Front and Back silence buffering with 50-ms buffers
Prevents the destabilization of the ADPCM decoder when two messages are played back to back. Once the arbitrary silences are trimmed from the message, this step places a standard 50-millisecond silence buffer on the beginning and end of every message.
Message level normalization
Converts all the messages in a single application to a single standard volume level, so that listeners hear every message at the same volume.

8 KHz resampling

Converts the 11.025 KHz sample rate of the master file to an 8 KHz sampling rate.

Equalization

Performs the frequency manipulation required to make high-quality message playback over telephone lines.
ADPCM compression for A-law playback and Intervoice formatting
Compresses the linear format to ADPCM to reduce the size of the message files on the Intervoice system. This step is where you specify whether a message is to be played as -law or A-law, or to be compressed to 32 KB.
Using the Batch Converter
To use the Batch Converter, follow these steps:
From the main Sonic Foundry Batch Converter window, select File > Open Script. Navigate to and open the Intervoice Example Batch Script, IVBExampleBatchScriptV6.bcs. The Sonic Foundry Batch Converter IVBExampleBatchScriptV6.bcs window appears in the Script tab view.
Do not select any of the options on the window. Select the directory where you want the output files saved. In the File format area, set the Format and Template.
Note If you change any selection, save the changes as
a new batch script and retain the original sample batch script.
To choose the files that are to be converted, go to the main Sonic Foundry Batch Converter window and select File > Add Media, or click Add Media on the toolbar. In the Add Media dialog box, select all the.wav files to be converted and click Open to continue The Files to Convert window tab view appears.

<source file spec>

Indicates the files to be renamed or copied (wildcards are allowed) and the location of the files. The command copies and renames all.ivc files in the current directory.

<target directory>

Indicates the directory in which the renamed copies are placed. The default is the current directory, if this parameter is not used. Example: The following sample CRVoice2 command copies all the.ivc files in the c:\audio directory that begin with the bankabc prefix and renames them to the appropriate m.### filenames. The copies are placed in the c:\audio\bankabc directory. [c:\audio]Crvoice2 bankabc*.ivc c:\audio\bankabc
Maintaining Audio Quality
This chapter explains the purpose of master voice files and stresses the importance of editing the files before compressing them. It also stresses the importance of retaining original master voice files.
Voice File Quality Issues, 7-2
Voice File Quality Issues
Master.wav voice files are intended to store messages in a high-quality format for editing and archiving original messages. To prevent loss of quality, do not edit and reuse compressed messages. After an Intervoice master.wav file has been compressed into the.ivc, m.###, or runtime.wav format, it must not be changed or edited. If you need to change a voice file, edit the Intervoice.wav master file and then compress the new edited file into the.ivc, m.###, or runtime.wav format. The Intervoice.wav master format is an 11.025 KHz (or 11,025 Hz), 16-bit per sample, linear format. It provides a high-quality method of voice storage for editing and manipulation before compression. All editing and volume control adjustments should be made on the 16-bit per sample master.wav file before compression into the.ivc or.wav runtime format. After all editing of the master.wav voice file is complete, the file should be processed using the Batch Converter. No editing of any type should be done on the compressed file.
Keep two versions of the master file: one original and one for editing. Do not discard or edit the original master.
Note that the original masters of the.wav voice files should be retained for future use. After a.wav voice file is compressed into the.ivc format, some quality is lost. Even though the Batch Converter and Sound Forge can convert the.ivc or runtime.wav files back to master.wav file size, a substantial loss of quality occurs in the new master.wav files, and even more quality is lost in the second compression step. For this reason, all re-editing should be done on the original 16-bit master.wav file.

Loss of Voice Quality

Severe loss of voice quality occurs if you convert Intervoice.ivc files to.wav files, edit those.wav files, and try to convert the.wav files back to.ivc files. To correctly edit files, save a copy of the original 16-bit master.wav file, edit the original 16-bit master.wav file, and then convert the newly edited.wav into.ivc and m.### (or 8-bit runtime.wav) for use on the Intervoice system.

In this appendix.

FAQs, A-2
Frequently Asked Questions
This appendix provides some frequently asked message recording questions and the answers to those questions. Users should consult this chapter before calling for technical support. This chapter will be updated as new questions are encountered. Contact Sound Forge technical support at:
http://mediasoftware.sonypictures.com/support. (Intervoice recommends contacting Sonic Foundry first for any technical support issues, as Intervoice does not have specific support personnel for Sound Forge.)
For Intervoice technical support, see Technical Support on page xv. For a list of message recording FAQs, visit the Intervoice Web site at www.intervoice.com/support/techinfo/voicerecording.
The following questions and answers pertain to recording messages.

Recording Steps

What are the main steps to record a voice message?
The primary steps in the recording process are as follows:
Set levels. Perform the noise floor test (see page 2-12). Record messages (Refer to the voice recording software). Apply the batch script (see page 5-8). Run the CRVoice2 utility (see page 6-4), for InterSoft IQTalk only. Move the files to the Intervoice system.

6 Defining DC Offset

What is the DC offset and what does it do?
DC offset is the average vertical offset from 0 dB of the recorded wave form. This is normally visible in the Sound Forge window only in silent sections of the wave form. Most sound cards have some DC offset when you are recording. Usually you cannot see this offset in the Sound Forge window unless you expand the vertical gain nearly to the limit and look into a silent portion of the recorded wave form. DC offset can cause problems in compressing voice files using the Intervoice ADPCM format. It also causes errors when you are trying to measure the noise floor of a recording. The Sound Forge Tools > Statistics command measures the average value or DC offset of a wave form along with several other important statistics about a wave form. This is the command used to check the noise floor of your recording setup. Since virtually all sound cards have some DC offset, it should be routinely removed as part of the standard processing of 11.025 KHz master files for compression into Intervoice formatted messages. A preset to remove DC offset should be placed in the standard Intervoice batch script that is applied to all Sound Forge voice messages prepared for runtime systems.

Installing the Audiophile 2496 Sound Card
When I installed the M-Audio Audiophile 2496 sound card, the software disabled the Windows volume controls on my system.
This is not an error or problem. The software is working correctly. Once the sound card is installed, the volume is controlled through the Midiman Audio Buddy preamplifier.

Turning Off AGC

How do I turn off the Automatic Gain Control (AGC) on my sound card?
The AGC that is on many sound cards must be turned off for any Intervoice voice message recordings. Unfortunately, Creative Labs turns on the Automatic Gain Control on all of their sound cards by default (except the AWE Gold). You can turn off AGC only by using the Creative mixer applet that comes with the Sound Blaster card in the application software package (ctmix32.exe). The standard Microsoft Windows 95 mixer applet (yellow speaker) does not have the ability to change this setting on Creative Labs cards. To turn off the AGC, follow these steps:
Install Creative Labs Audio Mixer applet (ctmix32.exe). Start ctmix32.exe. Right-click the mixer face. Select the Input/Output settings. Uncheck Automatic Gain Control.
You should not make recordings with the Sound Card AGC enabled. It can cause background noise as well as clicks and pops in the final recording.
Normalizing the Message Volume
When is the best time to normalize the volume so that it matches the messages already on the system?
The messages should be saved in 11, 025 KHz, 16-bit, monophonic, linear.wav format and recorded as loud as possible without clipping. All the subsequent processing can be performed using the Batch Converter. Generally, reducing the volume of the message is one of the last steps before compressing the.wav file into the runtime format.

Filtering Messages

When I am converting messages to the Intervoice ADPCM format, what filtering or optimization is necessary for the telephony network?
Converting the.wav Files to Runtime Format Overview on page 5-4 describes the steps for postprocessing a.wav file to InterSoft.ivc file format. The most critical step for voice clarity is equalization. By using the parametric equalization function in the Batch Converter to change the low frequency cutoff of the recorded voice, you can significantly change the clarity of the voice over the phone. Most users find that they can achieve their voice quality requirements by modifying the low-shelf frequency of the paragraphic EQ step. Usually they raise the low-shelf to between 200 and 300 Mhz for improved clarity. You may find that a modification of the high-shelf parameter also will improve some types of voices. If you want, you can boost a small portion of the audio spectrum centered at 800 Hz up a few decibels to get extra clarity voice. Do not improve the voice by boosting the bass. While this may seem improved on a hi-fi speaker, the voice quality does not work well over a phone handset. The primary information carrying bandwidth of the human voice is between 400 and 2000 Hz, and boosting frequencies outside of this range actually lowers the listeners comprehension.

Reducing Pops and Clicks in Messages
I have heard that the CRVoice2 utility fixes the pops and clicks in messages. Is this a result of making sure the message is in 4-byte increments, or does something else resolve this problem? And what does the CRVoice2 utility do?
Sound Forge version 4.5 and later versions of the Batch Converter with the Example Batch Script Vx.0.bcs file create InterSoft.ivc files with the correct byte count lengths. If you convert your.wav message masters using Sound Forge 4.5 or later versions of the Batch Converter, which is the correct way to convert messages, you do not have problems with pops caused by message length. Other wave editors do not process or compress the message files correctly and should not be used to create InterSoft.ivc files. Other editors can be used to record the.wav master files, but only the Sonic Foundry Batch Converter can convert them.
The CRVoice2 utility checks the message length and corrects it when necessary. However, it should not be used for that purpose because it does not perform the other operations on the.wav file that the Batch Converter does. The CRVoice2 utility is used strictly to rename the.ivc files to m.### for Intervoice systems. That is all it is designed to do. Checking the message length was included in this utility to fix a bug in an earlier version of Sound Forge. This earlier version is no longer available.

Glossary

-law (pronounced mu-law)
A Pulse Code Modulation (PCM) voice-encoding standard in which analog voice is band-limited to 4-KHz bandwidth and then sampled at twice its highest frequency (4000 x 2), with the sample then converted as an eight-bit binary number. Samples must be taken 8000 times per second, so each sample is 1/8000 of a second, or 125 microseconds. A standard telephone voice channel is transmitted at 64 KBps, or eight samples in 125 microseconds. The -law standard is used on T1 systems in the United States, Japan, Mexico, and Canada. See also A-law. A method of encoding digitized voice that requires fewer bits than linear or -law encoding (voice compression). The Intervoice system has a standard file format for voice messages using this compression method at 32 KBps and 24 KBps.
adaptive differential pulse code modulation (ADPCM)

See adaptive differential pulse code modulation.
The method of encoding sampled audio waveforms used in the 2048-KB, 30-channel PCM primary system. A-law is the standard widely used outside of North America on E1 systems. See also -law.

automatic labeling

The function of Sound Forge that allows the user to assign descriptive filenames to master voice files. This function should be set up before messages are recorded.

bandwidth

The range of electrical frequencies a device can employ.

Glossary-1

The Sonic Foundry program that allows the processing of multiple audio files in a single process.
A sound, or noise, that occurs in voice recordings; a type of audible interference that must be corrected in voice recordings.

command prompt

An operating system prompt that appears in an operating system window or full-screen session. Commands that run programs are entered at this prompt.

CRVoice2 utility

A utility that assists in the naming of multiple voice files for use in InterSoft runtime.
A decibel below 1 milliwatt (mW). See also decibel.

DC offset

The average of all of the positive and negative integers in a waveform. This number can be positive or negative. Zero is referred to as no DC offset.

decibel (dB)

A unit for measuring the power of a sound or the strength of a signal.

direct memory access

See DMA.
DMA (direct memory access)
A technique in which an adapter bypasses a computers CPU and handles the transfer of data between itself and the systems memory directly.

equalization

The frequency manipulation of a sound using adjustable filters.

filename

A name given by the user to electronically stored data for purposes of saving it and retrieving it at a later time. Filenames may be of varying lengths depending on the operating system used. The three-character extension (in the 8.3 format) designates the type of file: text (.txt), word processing (.doc), executable program (.exe), and so on. For multiple-operating system compatibility, the system requires the 8.3 format and recommends lowercase filenaming.

Glossary-2

imsgmod4.exe utility
A utility that truncates messages to a length divisible by 4 bytes. This truncation eliminates clicks and pops within messages used on the system. This utility is not recommended by Intervoice.

InterSoft system

An Intervoice system that operates on the InterSoft runtime environment with IQTalk or the VoiceXML Browser.

InterSoft version 3

The core software program used to bring applications into IQTalk voice browser runtime. InterSoft version 3 is designed to run on Windows NT and Windows 2000 operating systems. Intervoice IQ Environment [VoiceXML Browser System] A completely integrated environment that is bundled with the Intervoice InterSoft runtime system. However, it can also have speech recognition and text-to-speech servers. It offers a complete solution including design, development, and deployment tools as well as operation administration and management tools. It also includes data integration options to connect to legacy, enterprise server, and Web-based data sources. Browser options are IQTalk or VoiceXML, and operating system options are Windows NT or Windows 2000.

Intervoice runtime system
One or more Intervoice voice browser systems and other hardware that perform call processing, audiotex, speech recognition, call directing, voice messaging, and interactive voice response. This system is the interface between telephone callers and a host computer, the Internet, a local area network (LAN), a wide area network (WAN), or telecommunication network, and local or remote databases. The Intervoice voice browser systems can be integrated in one box or distributed in multiple boxes. This can also be called IVR and was formerly called Omvia Media Server.

IQTalk

The InterSoft runtime engine and voice browser that interprets and processes the InVision application. It can also interface with userdefined programs written in C or Assembly.
Kilobits per second. This is the number of bits passing a specific point per second in an audio file/voice message.

message

An audio file that, combined with other messages, comprises the spoken information in an application.

Glossary-3

Mu-law
See -law. A process used in the Intervoice Example Batch V2.sfz file to adjust the volume of all the audio files to the same level.

normalization

Omvia Media Server
See Intervoice runtime system.

preset

An audio processing function in Sound Forge.

RealCare

The remote diagnostic and troubleshooting service available to customers as a contractual option. RealCare operates a national WATS line at 1-800-955-4688.

runtime

The mode in which the system answers customer calls and communicates with the host; the system in the production environment.
A program, created by Sonic Foundry, and distributed by Sony Media Software, Inc., that performs postprocessing and complex editing. It can also be used to record and edit messages.

trimming

The action of eliminating or cutting silence from the beginnings and ends of voice messages.

voice compression

A method for reducing the bandwidth required for storing or transmitting digital voice signals.

voice file set

A set of individual voice files that contain the recorded messages for an application. These voice files are contained in a subdirectory of the application directory.

Glossary-4

voice message A digital.wav file used when an individual message is changed in a voice file set (.vfs). While the message file is being changed, any
application playing the current message continues to play it. The new message is played at the next message request.
voice message file numbering
The method of naming voice message files m.###. They must begin with m to be recognized, for example:
m.001 message number 1 m.777 message number 777 m2.010 message number 2010

Voice Pool Set

A type of VoiceSet that contains a collection of voice messages for applications that record large amounts of voice, such as LeaveWord recording. Messages from a Voice Pool Set (as in an InVision application) are used in place of messages from the voice file set (.vfs). Voice Pool Sets are contained in a subdirectory of the application directory. See also VSUTIL utility.

Voice Set utility

See VSUTIL utility.

voice talent

The person whose voice is used to produce voice recordings.
VSUTIL utility (VoiceSet utility)
The program that creates VoiceSet directories based on the information it reads in the.app file (stored in the application directory). Actual pools are created in the Voice Pool Set for use with LeaveWord and Advanced LeaveWord, as well as in the Alternate Message table for use by YourVoice. It also creates any necessary indirection files. VSUTIL should be started from a command prompt in the application directory.
A standard format for audio files. A file that contains Windows-compliant audio data. The file can be played on any Windows system with audio hardware. Intervoice has defined an 11.025 KHz, 16-bit.wav file as its master format.

Glossary-5

Windows 2000
A network operating system built from Microsoft Windows NT technology. Windows 2000 provides support for both peer-to-peer networking as a work group and centralized administration networking as a domain. Windows 2000 has two major versions: Windows 2000 Server and Windows 2000 Professional. For IVRs, Intervoice recommends Windows 2000 Server, which supports communication between a larger number of systems. Windows 2000 Professional can be used to run applications on business desktops and laptop computers.