| ||||||
|
| ||||||
|
Like many other Mac users, I have been keenly anticipating the release of MacSpeech's iListen dictation software for a long time. A very long time. The MacSpeech folks concede that about two months ago, they had a tough decision to make. Cut features in order to make the Christmas selling season, or delay introduction of iListen into next year. The company says that when its team we embarked on building iListen, they never dreamed it would take so long, and their investors (who took a considerable risk on a Mac-only company) were getting antsy to see results. Consequently, the current product is not what they originally envisioned. They chose their features carefully to emphasize their strongest points in version 1.0. MacSpeech engineers are hard at work on version 1.1, which has already entered testing. So, iListen 1.0.1 has been released in time for Christmas 2000, and MacSpeech commendably doesn't fudge the fact that it is still very much a work in progress, and a free update to iListen 1.1 will be offered to anyone who adopt early. My advice: go for it now. Even in its present iteration, iListen offers functionality that has never before being available on the Macintosh -- continuous speech recognition that can dictate into virtually any application or text field. Not only that, but the program seems remarkably stable for version 1 software, at least on my 233 MHz WallStreet PowerBook. iListen is not especially crash prone, and I have not yet experienced any loss of data, whereas its main competitor, IBM ViaVoice has eaten unsaved dictation more times than I like to recall (I am loath to use the ViaVoice save function because it has all the speed of cold molasses running uphill in the winter time. There are still some bugs to be worked out of iListen, but certainly no serious stability problems. One annoyance is that iListen will occasionally get stuck in dictation "pause" mode, and the only cure that I have discovered is to quit the program and restart it. I find that much preferable to Via Voice's sudden crashes with no warning, but it is still very irritating. Another problem that affects me, is that iListen apparently has a bad habit of breaking AppleScripts in Tex Edit Plus. I am not 100 percent sure that iListen is the culprit here, but the problem first showed up when I started using iListen. I have also noticed a general trend toward memory collisions and program freezes since I installed iListen, and I've been restarting a lot more often than I'm accustomed to as a consequence. I'm hoping that iListen will more peacefully coexist other applications in future versions. IBM beat MacSpeech to market with continuous speech dictation on the Mac by more than a year, but "taking the time to do it right" shows. The IBM product, which I have used extensively since last winter, is not a bad application by any means,. It has been a job-saver for me due to health problems that are making it increasingly difficult type more than a few lines. However, it is very much a port of a Windows application, with the essential and inescapable angularities and clunkiness that entails, while iListen is Mac from the get-go. Well, almost. iListen uses the FreeSpeech 2000 Speech Recognition engine from Philips Speech Processing, which was originally developed for Windows as well, so there is porting of technology involved, and consequent compromises. Nevertheless, iListen is at a much more Mac-like product than ViaVoice. Because many people considering iListen are already ViaVoice users, comparisons are inevitable, and MacSpeech does not shy away from them. The two products, while nominally doing the same thing, are actually quite different in many aspects. The most noticeable difference is that iListen will, as aforementioned, dictate into almost anywhere you can type, while ViaVoice requires you to dictate into either its own, dedicated SpeakPad word-processing window, or alternatively directly into a very short list of specifically supported applications. The advantage of iListen here is obvious. I have never used the ViaVoice Enhanced Edition with the direct dictation option. (For a review of VVEE visit here). That version does not support my 233 MHz PowerBook (a 300 MHz G3 is the minimum supported), and the handful of applications supported (AppleWorks, plus several Microsoft apps. and Netscape) are not ones that I use by preference. From what I hear, the ViaVoice direct dictation mode is pretty poky anyway, and you lose one advantage ViaVoice still has over iListen -- the text editing and correction functions in SpeakPad. There are also substantial distinctions between the two products in terms of system requirements. While ViaVoice runs comfortably on my PowerBook in its standard 12,012 k RAM partition (and I have not found that increasing its allotment to 30 or 50 MB speeds it up appreciably), iListen wants a whopping 60.8 megabytes of RAM. While MacSpeech says that iListen will technically work on a Mac with 96 megabytes of real RAM by setting the virtual memory at 129 MB (or using RAM Doubler), I would not recommend trying it without a minimum of 128 megabytes of real RAM aboard. I upgraded my RAM to 192 MB in anticipation of the iListen release, and find that works well for me, although it's still not any too much for the bunch of applications I run. If you have a RAM-challenged Mac (less than 128 MB), then ViaVoice, or the older, discreet speech, Voice Power Pro, are your only practical choices in dictation software. On the other hand, iListen is more modest in its processor demands than ViaVoice is, specifying a 233 MHz G3 as the minimum supported configuration, as opposed ViaVoice Enhanced Edition's 300 MHz G3. ViaVoice Millennium Edition, which I have, and which is still available, nominally also supports 233 MHz G3 machines, but in my experience, performance is not satisfactory. ViaVoice ME just crawls on my PowerBook. Dictation itself isn't too bad, but opening and closing windows, saving documents, and so on, are all excruciatingly slow (like about 20 seconds just to open a new SpeakPad window, and over a minute to save a document). Actually, iListen runs on any PCI Macintosh, and I have actually achieved decent performance with it on my 200 MHz 604e SuperMac S-900 (faster indeed than ViaVoice runs on my 233 MHz G3), but MacSpeech only officially supports Apple G3 installations. Some beta testers have also reported good results using iListen on older Macs upgraded to G3, but MacSpeech is not supporting those either, at least not for now. So what is iListen like to use? In a word, cool! The sense of freedom you get from being able to dictate into any program is wonderful, when you are accustomed to being cooped up in ViaVoice's SpeakPad. Installation is straightforward, but like all dictation programs, you will need to "train" iListen to recognize your voice before you can do anything useful with it. The training process is not much different than it is with ViaVoice -- ergo: you read the program some stories and it processes the data into personal voice profile files. I would rate the iListen training process as the better of the two, as it gives you a selection of 12 stories to read instead of five, and allows you to skip a problematic word if the training program gets "stuck." The only "problem" I had was that two of the iListen stories involve delicious-sounding Italian food, and reading them gave me a case of the munchies. On the downside, I would say that iListen needs more training than ViaVoice does. I got decent performance out of ViaVoice after reading it just one story, better with two, and I never bothered reading it more than three. I have read six so far to iListen, and it still appears to need more training, especially for single word and short phrase accuracy. Training either program takes some time, and then some more time while your computer processes the voice profile files. Once that chore is out of the way, you are ready to dictate.
Unlike ViaVoice, iListen has a floating control and feedback palette, which contains a microphone toggling button, a not terribly useful button that brings the iListen application front most, a mike input level meter bar, a text field that tells you what mode you're in (EG: idle; dictating; spelling). and a "pop-down" field that displays what the program thinks it heard you say. I found having the latter bouncing up and down distracting, and have turned it off in the Preferences. There is also an animated picture of a dog named Lykke who belongs to MacSpeech's CEO, and who wags her head and opens and closes her mouth somewhat pointlessly. Not being a dog fancier, I am gratified that several alternate animations are provided in the Preferences. I chose a grey and white cartoon cat, which more suits my tastes in quadrupeds, but I could happily get along with no animation at all in exchange for making the palette smaller, and retrieving whatever amount of processor power the animation absorbs. I encourage the good folks at MacSpeech to consider at least making the animation optional. When you start up iListen, the program takes a minute or so to load your voice profiles, and then you're ready to dictate, virtually anywhere. iListen has three modes: Dictation,Spelling,and Command. To switch among them, say “Switch to Command Mode” or “Switch to Dictation Mode.” You can also press the F11 key to rotate between modes. Pressing the F11 key will turn the microphone on and enter Dictation mode. While in Command Mode, you can say “Open” followed by the name of your favorite applications.
Tex Edit Plus is my default venue for dictating, and its standard features plus the F-Key toggled AppleScript macros I have added for text editing. help compensate for the current lack of voice-activated text editing and correction features in iListen 1.0.1. This lack is to be addressed in iListen 1.1. Here are my impressions of iListen's stronger and weaker points compared with ViaVoice: • ViaVoice recognizes both pronunciations of 'the' (iListen only recognizes 'thee' but not 'thuh'). • ViaVoice is much better with proper names, numbers, computer terms, and Canadian place names. ViaVoice also does a better job of spelling things like acronyms in the normal dictation mode, and handles money amounts more gracefully as well. iListen counters with its spelling mode, which is useful to have, but which I find rather slow and cumbersome, as well as far from dead accurate. For instance, it has a terrible time confusing b and d. • ViaVoice seems to have a significantly larger usable vocabulary, at least of the sort of words that I use frequently. ViaVoice for Mac includes a 64,000 active word vocabulary, while iListen has a native vocabulary of only about 30,000 words and a background vocabulary of 300,000 words developed in partnership with Oxford University Press. ViaVoice is definitely better-equipped here in practical terms. ViaVoice also allows you to add more words to your dictionary, which iListen currently does not. • iListen handles punctuation much more gracefully than ViaVoice. • iListen has a disconcerting tendency to ignore or drop words. ViaVoice rarely does this. • iListen's speech engine seems to be much more context-dependent than ViaVoice's. When dictating long passages without interruptions, I find that iListen has substantially better accuracy than ViaVoice, so long as it knows the words. However, because of the smaller dictionary mentioned above, it does stumble over more specific words than ViaVoice does. • ViaVoice is definitely happier dictating a single word or short phrase then iListen is, which is unfortunate for iListen since one of its strongest points is the ability to dictate into text fields like save boxes or Web browser forms where single words and short phrases are the likeliest thing to be dictated. • iListen is much more tolerant of incidental noise in the background then is ViaVoice (same microphone and location with both programs). • I really miss the editing commands available in ViaVoice when using iListen, but that is hardly a fair criticism of iListen because the editing only works in ViaVoice when you are dictating into the dedicated SpeakPad window, and not when you are dictating directly into the few dedicated applications that ViaVoice Enhanced Edition supports. Personally, I don't miss correction window in ViaVoice a whole lot because it runs so slowly on my 233 MHz PowerBook that I've virtually never used it, but I love that voice activated editing. As noted above, MacSpeech will be offering all registered users of iListen 1.0 a free upgrade to version 1.1 when it becomes available within the next 1-2 months. Registered users will be able to download the upgrade at no cost, or receive a new CD and user ’s manual for the cost of shipping and handling. iListen 1.1 will incorporate correction, which will allow the user to move forward and backward in the document using their voice, as well as provide alternatives for misrecognized words. The user’s voice model will be automatically adjusted when corrections are made using iListen 1.1 ’s interface. Also incorporated in iListen 1.1 will be: 'Learn My Writing Style,' comparable to the 'Analyze My Documents' feature in Via Voice Enhanced Edition.The user can select multiple documents from their hard drive and drag them onto the Learn My Writing Style window within iListen 1.1 (you cannot do this with Via Voice). Also, iListen 1.1 has the MacLink Plus translators built-in so the user can select documents from virtually any application that supports text (Via Voice only analyzes plain text documents). We will have to wait for iListen 1.2 before the 'Learn One Word' feature is incorporated to train iListen to understand words that are not already in either its native or background vocabularies. iListen 1.1 or 1.2 should be fully carbon compliant. MacSpeech plans to ship a version of iListen for Mac OS X by the time Mac OS X 1.0 is released to the public. To summarize, notwithstanding the criticisms I've cited in this review, many of which will be rectified next two version upgrades of the program, I am powerfully impressed with iListen. I had developed a functional working relationship with ViaVoice, and would probably continue to use it for some tasks, except that the MacSpeech extensions seem to conflict with the ViaVoice extension, so switching between the programs involves mucking around in Extensions Manager, and then restarting the Mac -- too much trouble to be practical. The one I will be using from now on will be mostly iListen. I guess that about says it all. iListen 1.0 is being sold for the introductory price of $99 without a headset, including a free upgrade to version 1.1. The software is exclusively from MacSpeech's web site. It can be downloaded immediately. A boxed version will begin shipping in mid-December. For more information, visit:
Appendix Key Features and Benefits
|
. |
Cool Mac Gear• iPod 1G-2G • iPod 3G • iPod 4G • iPod Mini • PowerBook-iBook • Keyboard Skins • Garageband
| ||||
Most of them will automatically open. You can also voice activate various finder functions and keystrokes using command mode, including mouse clicks, page up and page down, and the home and end keys, among others. Adding a word to the Dictation Commands vocabulary makes it available when in Dictation Mode. Add things like names, or frequently used commands such as “Tab Key” to this command set. You can also write AppleScripts to function as voice-activated macros.