Speech-to-text software, sometimes known as dictation software, is something that lets you talk to the computer in some form and have the computer react appropriately to what you are saying. This is totally different to text-to-speech software, which is software can read out text already in the computer.

Command and Control Software

There are two types of speech-to-text software available. One type is called "command and control" and it lets you speak commands to your computer to control it; hence the name. For example, a command that the computer understands might be, "go to the Apple website" or, "tell me the time". Each command is pre-programmed and the computer will only recognise those commands it's been programmed for; you can't use this software to write an email or use iChat for example.

Command and control software for the Mac - known as "Speakable Items" (or sometimes, confusingly, "spoken commands") - is already built into every OS X computer, although most people don't know about it. You don't need to download, buy, or install anything to get this software to work, just a microphone that works with your computer. The main drawback is that the Speakable Items software programmed for English with a standard American accent, and has significant trouble with any other accent. It doesn't function at all with languages other than English.

Some resources for getting you up and running with Speakable Items include:

• Get Siri-like Functionality on your Mac with Speakable Items (video with full transcript available)
• Apple Support Articles about Speakable Items

Dictation Software

The other type of speech-to-text software is usually called "dictation" software. This is the type that lets you write an article like this one, type stuff to your friends in iChat, or type an email. The most common Windows software for speech to text dictation - you've probably heard of it - is Dragon NaturallySpeaking. There is only one dictation-capable speech-to-text software available for OS X which is being updated and developed and it's [msd]. Dictate is the successor to a program named iListen which MacSpeech used to produce.

Like all dictation-capable text-to-speech products, MacSpeech Dictate works very well for some people and very badly for others. Whether it will work for you depends on many things including: how much effort you're willing to put into learning it, how good your microphone is, your age (text to speech usually works less well for children), how much your accent matches what the program expects, and whether your voice changes a lot through the day.

MacSpeech Dictate is also still fairly new software - it was only released on the 15th of February, 2008. In comparison, the premiere speech recognition program for Windows is Dragon NaturallySpeaking which has been in development since the 1980s[1].

When MacSpeech Dictate was originally released it had several major problems which made it unusable for people with disabilities, but most of these have now been resolved:

• There was no good help functions inside the application - this was rectified in Dictate version 1.3
• It didn't learn from corrections - this was rectified in Dictate version 1.2
• Couldn't spell words out by voice - this was rectified in Dictate version 1.2
• Couldn't request individual key presses (such as command-s or command-option-escape) by voice - this was rectified in Dictate version 1.3
• Couldn't be taught new words, such as names or jargon specific to your profession - this was largely rectified in Dictate version 1.2, although some words still resist training
• There was no way to control the mouse by voice - this was finally rectified in Dictate version 2.0.

I tried using the old iListen program a few years ago and could not get results that were useful, an on-screen keyboard was the best solution at the time. Although MacSpeech Dictate is in its early days as a program, its recognition of my particular voice is hugely better than iListen's was. This is not surprising though, as MacSpeech Dictate's speech recognition engine is based on the same engine used by Windows' Dragon NaturallySpeaking - widely recognised as the best consumer speech recognition available.

[msd] requires the requires Intel-based Macintosh hardware and requires Mac OS X 10.5.6 (Leopard) and higher. Thirteen English dialects/accents are supported, and US and UK spelling options. These are:

• US Spelling
  • American
  • American - Inland Northern
  • American - Southern
  • American - Teens
  • Australian
  • British
  • Indian
  • Latino
  • Southeast Asian
• UK Spelling
  • Australian
  
  
  • British
  
  
  • Indian
  
  
  • Southeast Asian

Specialised versions - Dictate Medical and Dictate Legal - are available for dictating in these language areas, and Dictate International is now available and recognises speech in French, German, and Italian. MacSpeech have strongly hinted that Spanish language recognition is next on their agenda.

MacSpeech Dictate is a great program for dictation and some computer control, but it is not something that will let you control the computer completely "hands free". For quadriplegic users and others who need full computer control, you will need to supplement Dictate with use of a mouth stick and keyboard, or a program such as SwitchXS for switch access to functions not available by voice. I highly recommend Dictate though, it's part of my suite of accessibility technology and I use it whenever I am able to.

Website: [msd]

- Ricky Buchanan

[msddisclaim]

[msdbanner]

It's been announced that the MacSpeech company has been purchased by Nuance. Nuance are the company behind the Windows product "Dragon NaturallySpeaking" and other recent Dragon products for iPod Touch and iPhone.

So what does this mean for [msd] and the other MacSpeech products? Nuance are quick to assure us that "nothing will change in the near term" but I think things will change for the better. Nuance is a much bigger company than MacSpeech - the Windows market has always been more than ten times bigger than the Mac OS X market, and the company which is now Nuance has been around much longer than MacSpeech has. I don't know the number of employees that either Nuance or MacSpeech actually has, but I'd be willing to bet that Nuance has a lot more. And this will probably mean good things for MacSpeech's products, as more talent is available to work on them they can progress more quickly.

MacSpeech Dictate still has some major features missing, such as mouse control, but it's growing and maturing quickly. One big problem for people wanting to switch from using Windows with NaturallySpeaking to using OS X with Dictate is that the names of various commands are completely different between the two products. Actually, Dictate's command set has grown in a haphazard way and commands are difficult to memorise because different commands are constructed in different ways. I would think that with the acquisition of MacSpeech, there's the possibility of the Dictate command set becoming more like the NaturallySpeaking command set. This may be confusing for existing customers, but it would be a huge blessing for customers switching from Windows to Mac and, I think, in the long run it would be a good thing.

Nuance obviously thinks the OS X market for voice recognition is growing and viable, and they're willing to spend money to get into it. This might mean that MacSpeech products eventually cost the same as the equivalent NaturallySpeaking products - at the moment the Mac versions cost significantly more, despite having fewer features. Reducing the cost would not have been possible for MacSpeech alone, as they needed cash flow, but Nuance are in a stronger financial position and they're bigger so they can probably cope with a bumpier cash flow. Cheaper assistive technology is good for everybody, so I hope this one comes true!

I've got a MacSpeech Dictate-related free download coming up soon too - so regular readers stay tuned!

What else do you think will change, or hope will change, with Nuance's buyout of MacSpeech?

- Ricky Buchanan

[msdbanner]

[msddisclaim]

One of the major things that the MacSpeech Dictate family has been lacking is the ability to take pre-recorded files and convert them to text. Not any more: MacSpeech Scribe will do just that for you, with up to 99% accuracy.

MacSpeech Scribe will accept any file in one of these formats:

• .wav
• .aif or .aiff
• .m4v, .mp4, or .m4a

Audio file quality will affect the quality of your recognition, of course, so using a certified recording device is recommended, but not required - anything that will produce the correct file format will work. At the moment, the iPhone, iPod Touch, and several Olympus digital voice recorders are the only devices certified but I would expect that MacSpeech expands this range fairly quickly.

Recording a sound file to run through Scribe is pretty much like using MacSpeech Dictate itself, but without the ability to correct and train phrases as you go. If you want your transcribed document to include punctuation, you need to speak the punctuation signs into the recording, and you need to train MacSpeech Scribe to the voice of the person who recorded the audio file before it can transcribe.

So what are the limitations? Bear in mind that I have not had access to MacSpeech Scribe myself, but these are the limits that have been described by MacSpeech or can be inferred from the behaviour of other products in the MacSpeech family:

• You can only have one speaker per file, so MacSpeech Scribe will not be helpful for transcribing a meeting or class or any other situation where there is more than one speaker.
• The program must be trained to the voice in the recording, so it's also unlikely to be useful for transcribing a speech or lecture unless the speaker is willing to spend some time with you creating a profile for MacSpeech Scribe.
• Because of the need for punctuation to be spoken aloud, I am not sure if the accuracy would be adequate in a situation where punctuation was not spoken - from Scribe's perspective the text produced would be one really long paragraph.
• We know from other MacSpeech products that the distance from mouth to microphone is very important for recognition, so I would think any speaker who is moving around would significantly degrade accuracy. If you need to record a speaker like this for MacSpeech Scribe's use I would suggest investing in a lapel microphone for your recorder.
• Background noise or any other non-speech noise in your recording will also degrade accuracy. Get a directional microphone for your voice recorder so it only picks up your own voice, or dictate in a quiet place.
• Changes in voice quality from emotion or emphasis also degrade recognition. [msd], in my experience, does best with a very steady tone of voice - not a monotone but no getting excited or sad or speaking too fast or too slowly - so I would expect that MacSpeech Scribe is similar in this respect.

MacSpeech quotes that:

MacSpeech Scribe lets you easily add new words and acronyms, edit and navigate transcribed documents, and so much more. MacSpeech Scribe makes it easy to work with your transcribed document so you can create the perfect document for your needs.

which leaves me unsure if its editing abilities are the same as other MacSpeech products and, if they are, does it let you verbally add a word or phrase that was missed by the dictation engine? If so, what does Scribe not have that Dictate has? I'll have to get hold of it to clarify that one for you!

MacSpeech Scribe is available immediately, in English only, for all the dialects of English usually recognised by MacSpeech products. There is a special price of US$99 for currently registered MacSpeech Dictate 1.5 customers, the regularly suggested retail priced is US$149.

- Ricky Buchanan

[msdbanner]

[msddisclaim]

Photo credit to Twon.

Bakari, of the Mac Photography Tips blog, has made a great tutorial video which shows one way to create new commands for MacSpeech Dictate - using the "Menu Item" command type.

There are other types of MacSpeech Dictate commands which you can make that will let you do things that don't have a menu item, but this is one very simple way to create a new command without needing to know anything about programming or other complicated "geek stuff". I suggest you watch the movie in full screen mode so you can see what Bakari's doing:

[embed width="640" height="385"]http://www.youtube.com/watch?v=iaE4VzWUxZ8[/embed]

At the end of the video Bakari says that one of the limitations is that MacSpeech Dictate can't push the "Post" button in the Tweetie program because it has no menu item. However, as he hovered the mouse over that button I clearly saw the tooltip showing "⌘⏎" which are the symbols for the comand and enter keys. MacSpeech Dictate can do keystroke combinations with the verbal command "Press the keys..." or "Press the key combo..." so if we tell MacSpeech Dictate "Press the keys command enter" or, "Press the key combo command enter" then the tweet will be sent!

It's not as neat as making a new command named "send tweet" or similar, but it works. Note that a new AppleScript command could also be used, but this is a bit more "geeky" and requires a touch of programming, so I won't go into it here.

A note: I found that to use the "press the key combo" command that it's important to say the whole command as if it's one phrase. My instinct is to say the command as two phrases, as if the punctuation were something like this:

Press the key combo, "Command Enter".

But this makes MacSpeech fail to recognise the command because I'm pronouncing it as two phrases where the program expects one. If I say it all as one phrase, more like this:

Press the key combo command enter.

Then the recognition is very good.

Do you have any programs you've made MacSpeech Dictate commands for, or that you'd like to make commands for and don't know how? Leave a comment and I might write about commands for your specific needs in future articles!

- Ricky Buchanan

When trying to dictate using MacSpeech, things can be very frustrating at first. Trying to learn to use a new program and memorise many commands at the same time as you are learning the art of dictation is always going to be difficult. Happily, there is one thing which you can do that will greatly increase the chances that you will learn to use [msd] usefully.

Select 'Preferences' from the Dictate menu to open MacSpeech Dictate's preference pane. Click on the recognition icon and check the box labelled "Always show recognition window when dictating". Make sure the other checkbox, the one labelled "Close recognition window after each choice", is unchecked as shown here:

You may also wish to increase the maximum number of alternatives shown in the recognition window, and if you have a fast computer then moving the slider towards the end marked "Accuracy" may also improve your recognition.

Having the recognition window open whenever you are dictating will make it much more likely that you will use your voice for corrections. This will, in turn, allow MacSpeech dictate to fine tune your voice profile each time you use the program. Having a better voice profile will reduce the number of correction which are needed as Dictate will recognise your voice more accurately.

- Ricky Buchanan

[msddisclaim]

[msdbanner]