field-theory.org
the blog of wolfram schroers

Really virtual — Dragon NaturallySpeaking on MacOS X

I describe how to install and setup Dragon NaturallySpeaking (Dragon) on MacOS X via Parallels Desktop running a Windows 7 virtual machine. This configuration is not officially supported by Nuance, but it is an ideal solution for Mac users who want or need to run the Windows version of Dragon, but still want the full power and convenience of MacOS at the same time.

Table of contents:

Introduction
Installation and setup
Usage and work-flow
Summary

Introduction

This article describes how to use Dragon NaturallySpeaking on a virtual machine with Windows 7 on an Apple machine with MacOS X using Parallels Desktop. Such a setup is not unlike a remote virtual environment like Citrix.

My specific setup at the time this article was written (February 2011) is:

Both the successful configuration and potential problems and pitfalls are discussed.

Installation and setup

Installation of Parallels Desktop, Microsoft Windows and Dragon NaturallySpeaking is explained in the respective user manuals and I will not go into details here. The point where we need to take special action is when a speech profile is first created.

First, the Windows virtual machine should get the maximum resources it can. Referring to the Parallels manual make sure to select “Faster virtual machine” in the “Performance” menu and turn “Tune Windows for speed” on. In reference to the General settings it is strongly advisable to grant Windows as much memory as recommended, but not more (this will depend on the amount of physical memory installed in your machine). Personally, I do not increase the number of CPUs (although I have a dual-core Intel laptop) as Dragon only minimally benefits from multiple CPUs. However, if you have more than two cores, it won't harm to give Windows two cores.

The important point is to use a USB headset and register the headset with the Windows operating system, not with MacOS. For further details, please refer to the user manual of Parallels Desktop. If you are running this setup for the first time, simply selecting “Windows” while holding the Alt-key pressed should do the trick. Otherwise, you may need to use the USB-Preferences dialog as explained in the user manual.

Now you can select “Profile” and “New User Profile …” in the Dragon menubar. Make sure that you select “USB” as the “Microphone” source:

Selecting USB headset

You will later be asked for the audio source you want to use for dictation. Make sure that you select your USB headset and stay clear of the “Parallels Audio Controller”:

Choosing the USB headset

Now proceed with the initial setup and training of the profile as usual.

Note that I always click on the “Play” in this dialog to verify that my audio system is working properly:

Verifying the audio setup

While you are dictating text you have one more option to verify your audio settings. If you think that the recognition accuracy is surprisingly low then dictate a phrase of text, select it and then either use the voice command “play that back” or press the play button on the Dragon Bar:

Playing back text you have dictated

If you hear any distortions or missing pieces of sound, your headset may not be working as a USB device. Please see above how to ascertain that it does!

IMPORTANT!

I do not recommend to use an analog headset! Similarly, I do not recommend to connect the USB headset to MacOS instead of the virtual machine. There are two reasons:

  1. The Dragon software usually complains if you plug in the “wrong” headset. In this situation it will not complain! It is even worse if you accidentally forward your laptop's built-in or secondary microphone: The only thing you will notice is that the recognition accuracy suddenly drops from 99% to 55% and you stand there wondering why! Ouch!
  2. When the host starts to swap out memory, audio information may get dropped. The sound will sound choppy or get artifacts that will reduce recognition accuracy or that may even drop entire phrases.

Both situations are avoided by using a USB headset and connecting the headset properly to the virtual machine as described above.

Note that it is not possible to directly dictate into Mac applications. If you rely on this feature the best solution is the MacOS-native software Dragon Dictate for Mac.

In theory it is possible to use a Virtual Network Computing (VNC) connection hosted on the Mac side. Apple has its own VNC-host called Apple Remote Desktop. In practice I found this approach too much hassle for too little benefit (it loses all capabilities to mix speech and keyboard input) and prefer to rather copy&paste information back and forth, very similar to Dragon's dictation box. For this reason I do not describe this setup in this article.

Usage and work-flow

Using the software is straightforward: First, you launch the virtual machine with Windows from Parallels, then start Dragon NaturallySpeaking within Windows. Windows can be run in any of the View Modes.

If the USB headset is not plugged in and configured correctly, Dragon will complain about it when the profile is loaded:

Error message when USB headset is missing

This is a safety precaution and one of the reasons why I recommended using a USB headset for the virtual machine.

At this point you can fully use Dragon and interact with all Windows programs just like in a native Windows environment. What I like to do is launch WordPad (voice command “open wordpad”), dictate my text and then use “select all” and “copy selection” to copy the text into the clipboard.

Copying text to clipboard

Using Command-Tab I switch to either TextEdit, Pages or Mail (the programs I commonly use dictation for) and press Command-V to paste in the text. This is very convenient and straightforward.

Pasting the text in a Mac application

The other way works fine, too. As WordPad is fully speech-aware, you can navigate and edit text copied from MacOS just like any other text you have typed in this application.

Summary

In the current form, digital speech recognition interacts well with both the virtual and the host operating system. Thus, it is possible to seamlessly copy and paste text between the two operating systems. At this time it is not possible to dictate directly into Apple applications. If you rely on this feature, you should check out Dragon Dictate for Mac.

The configuration discussed in this article is not officially supported by Nuance, but customization and support is available through resellers and specialized consultants like me.