Other Sites

MacSpeech's dictation applications for the Mac OS have had a long gestation period, and it's not quite over yet, but at MacWorld Expo MacSpeech announced that it would distribute 10,000 CD-ROMs containing a preview release of iListen, the company's high-end dictation product.

Chief MacSpeech evangelist Chuck Rogers kindly sent me one of the preview CDs to sample. Releasing preview versions of not-quite-ready software can be perilous, as Netscape discovered with Netscape 6, and Chuck tacitly confirmed this in asking me to "be kind."

He needn't have worried in terms of my friendly inclination. I'm a bit of a cheerleader for MacSpeech, which is run by a bunch of Mac people who have never wavered in their faithfulness to the platform. MacSpeech will be a Mac-only product, not a port of a Windows dictation application.

MacSpeech is headed by a team of software engineers led by company founder and CEO Andrew Taylor, some of whom were key members of the Dragon/Articulate Systems' PowerSecretary and Voice Navigator development teams.

MacSpeech's original mission statement was music to a Machead's ears:

"We are Macintosh people, committed to the Macintosh platform. We have worked with speech in the past and are working with it now. Speech on the Macintosh is our one and only focus! No other company can say this and no other company can offer you our focus, our desire and our will to succeed.

• We believe in Speech and we believe in Macintosh.

• We believe in the concept of Insanely Great.
• We believe in putting the user first and the computer last, possibly even behind that potted plant in the corner.
• We believe in getting things done and doing them right.
• We send neither allegiance nor tribute to N47 W122.

The MacSpeech folks note that being Mac-only means their engineers "don’t have their hands and creativity tied by a corporate requirement to maintain a common code-base across platforms. Since we are uniquely Macintosh-based, we can take advantage of all Macintosh has to offer."

The first dictation application for the Mac was Articulate Systems Power Secretary, which was absorbed by Dragon Systems, then sold to the British developer GT Interactive Europe, which continues to market it as Voice Power Pro 3.

Power Secretary/VP pro 3 is a capable application, allowing the user to dictate into almost any application or text field, and supporting menu and command control of the Mac, but its shortcoming is that it is not continuous speech capable. You have to pause between each word. On the other hand, if you have a less than state of the art Mac -- i.e.: anything slower than a 300 MHz G3 -- Voice Power Pro 3 may be your best or only practical choice in a dictation application, since the discreet speech engine makes much lighter demands on your system then does continuous speech.

The first continuous speech dictation product for the Mac was IBM's ViaVoice Millennium Edition. I use ViaVoice regularly, and the continuous speech feature is great, or at least it would be if I had a faster Mac. On my 233 MHz G3 series PowerBook, ViaVoice runs so slowly and sluggishly that there is really not any advantage in dictation speed over Voice Power Pro. In fact, I think VP Pro is actually faster on my PowerBook, although I have not done a timed head-to-head comparison. The problem is that while ViaVoice will accept continuous speech, my "slow" 233 MHz G3 processor can't keep up, even when I speak very slowly.

The other deficiency of ViaVoice is that except for a handful of specified applications, none of which I use, you must dictate into a dedicated window called SpeakPad, and then transfer your text via the clipboard to its intended destination. You can't dictate into Finder text fields, spread sheet cells, and so on, and you can't control Finder functions with voice.

Consequently, it would seem that a melding of Voice Power Pro's versatility and ViaVoice's continuous speech recognition would result in the ideal Mac dictation application,

iListen will be MacSpeech's high end dictation product, slated to incorporate dictation, editing, formatting and simple speech navigation with any application. You will be able to dictate text, edit and format it with just the power of your voice. iListen supports continuous speech recognition using the Philips FreeSpeech 2000 speech engine, and allows you to open, close and control applications and navigate the desktop by speech.

Command & Control is disabled for the Preview Release of iListen. However, to control your Macintosh with your voice you can still use MacSpeech's free PlainTalk enhancement, ListenDo!, which is included on the Preview Release CD. The commands "Wake Up" and "Sleep" also do not work in the Preview release, and correction is disabled.

The minimum system requirements cited for the eyeless and preview are:
• RAM 96 MB minimum, 128 MB or more recommended
• Processor PPC 604 233 MHz minimum ÃÂ? 300 MHz G3 or better recommended
• Disk Space 200Mb, including user files created by iListen
• System MacOS 9.0 recommended, MacOS 9.0.4 required for USB microphone.

OK, I thought, at least this time I don't have the absolute bottom of the pile machine as I do for ViaVoice's minimum system requirements. My PowerBook G3 should have a fair bit more processing grunt then a 604e of similar clock speed.

I ran the installer, a process that seemed to go smoothly enough, and then proceeded to go through the initial microphone setup and voice training exercises whereby you read specified text into the microphone so that the program can plot your voice patterns. The iListen training was not very different from that of ViaVoice, and took about 20 minutes while the prompt Windows walked me through the steps. Then I had to wait for a while during which the program manufactured a personalized voice profile.

When that was completed, I restarted my machine, and that was when I began encountering problems. As soon as I tried to dictate, the program would crash. After several tries, I decided to increase the program's RAM partition, but when I opened the get info box, I discovered that the minimum RAM allocation for iListen is 71,370 MB!

My PowerBook has 96 MB of RAM installed, and I usually run with that doubled using RAM Doubler. However, Chuck Rogers told me that iListen does not coexist happily with Virtual Memory in some cases, and I figured that it might be having similar problems with RAM Doubler. Now, as noted, the minimum system requirements for iListen say that 96 MB is sufficient, however, with Virtual Memory and RAM Doubler turned off, I found that I didn't have enough RAM to start up iListen, unless I turned off all but the barest minimum of control panels and extensions, which did allow iListen to start up, but then there was no extra RAM for running other applications, and indeed some of the MacOS functions that I depend on had also been disabled in the RAM economy exercise. Then I began getting system errors when I tried to run iListen.

At that point, I threw in the towel. My machine obviously does not have the muscle to run the Preview version of iListen satisfactorily. I did get it to dictate a few words between crashes, and my overall impression is that this will be a really good dictation program for people who have, say, a minimum of a 300 MHz G3 and at least 128 MB of RAM, although I would suggest even more RAM than that if you want to run more than a handful of applications simultaneously.

Indeed, MacSpeech has posted the following note on their Web page regarding minimum system requirements for the preview release:

"The Preview Release of iListen has greater minimum requirements than the shipping version will require. The Preview Release will work best on a G3/300 or better, with greater than 96MB of physical RAM and in most cases we recommend that Virtual Memory be turned off."

Chuck Rogers says that iListen PR1 has done extremely well in the field, but that 233 Mhz G3 PowerBook is one of the configurations that displayed the problems I experienced.

Chuck further notes:

"The RAM footprint for something that has five continuously operating recognizers is enormous. You must remember that we are doing Command & Control, Spelling, and Dictation, as well as polling for PlainTalk usage.

"There is also a considerable hit to RAM as a sacrifice for talking into any application. If all we did is talk into our own, we could reduce the RAM footprint considerably -- even Via Voice enhanced doesn't REALLY talk directly into the 'supported' applications. It talks into a background app and then translates text via AppleEvents. iListen really does pass the text directly to the application in pretty much the same way the keyboard does.

"The RAM requirements will lessen somewhat about two or three testing releases from now when we remove the legacy PlainTalk dependencies. Right now, the code transferred from ListenDo! is pretty much there in its entirety, which means PlainTalk (English Speech Recognition) has to be installed for iListen to work. This is on the list to get removed..., but I don't know if it will make it into the next public release.

"Short story is, this will reduce your RAM requirements providing you are willing to get rid of English Speech Recognition from your system.

"On the bright side, we will be coming out with iDictate later this year. Since we won't be doing Command & Control or working with PlainTalk with that product, we can get rid of two recognizers.

"So the bottom line is, you probably do need to buy more RAM if you want to do Command & Control."

I agree. My usual practice is to have somewhere between 15 and 20 applications open at once, and ViaVoice, works happily in this environment with a 30 MB RAM partition (the specified minimum use 10 MB) and RAM Doubler enabled. Of course, ViaVoice will not do a lot of the cool things that iListen can, which is presumably why it is a lot less RAM hungry.

I look forward to trying iListen on a more powerful Mac, and also to checking out its more basic sibling, iDictate, when the latter becomes available. I am convinced that these MacSpeech applications are destined to become the class of the field in Mac OS dictation software.

However, for those of us with older, slower Macs, I think that the best choice in dictation software will continue to be the venerable, but thoroughly developed and less hardware-demanding, Voice Power Pro.

The free CD (shipping charges apply) with the preview version of iListen is offered on the MacSpeech Website, and also includes software from 10 developer partners who have joined forces with MacSpeech to offer demos and special offers on their products. For example Nisus Writer 4.1.6; a demo of Nisus Writer 5.1.3; and Nisus Email 1.0.9 are included.

iListen will be available for purchase as soon as final testing is complete. It will sell for a suggested retail price of $169, including either a regular audio or USB headset microphone. The software, without a microphone, will be available for $129.

For more information, visit:
http://www.macspeech.com