To enable speech recognition on windows 2000 and maybe others...

Summary

You download some files from Microsoft, then you install them. It's not shareware, it's not adware, it's totally free! Then you find the sample application called DictPad in the installation folder system and run it. It is where you will be doing most of your dictation.

If you discover something that's not on this page, as I'm sure you will, that would help people, use the form at the bottom to drop me a line. Go There.

Notes

It uses the Microsoft Speech Engine and what I think is an ActiveX control. it's a very large file to download (the big one is 68 MB), but for me it's worth it. If you go to the MS page for the larger file, MS claims they willsend you a CD. It may even already be on XP.

after twenty minutes of training the machine (A. K. A. reading to it), its accuracy is very good. I'm even putting in the symbols/punctuation by voice.

There are very few help files available with this, and even the ones included are not very good (unless you are a programmer), but this page should contain enough documentation to get you up & running. Maybe this is why it's not well known.

I guess one reason this is not very popular is that it's part of the Microsoft software development kit for the a speech engine. You get all the extra stuff as if you were a developer, but you probably won't need it. its also possible that i'll figure out exactly which files that you need, and then might be able to zip together a condensed version much smaller than the original. but I'm pretty busy these days , so don't count on it. ;-)

Currently it will not type directly into Word or any other programs, but I'm really going to enjoy using it for e-mails and recording personal documentation on the fly. All I had to do is open a small application (DictPad), click a button, and speak into the microphone! I can edit it here, or wait and edit after I paste the text into another application. I think one of the users below said that it would work in Word, but I haven't gotten that far, nor seen how it is easily done.

I'm not even using a fancy microphone. I bought it for $10.00 at Walmart . Though I'm sure that an expensive one that is a headset/boom mic would allow even more accuracy, much less comfort as you speak to it.

It's also very good at recording your side of telephone conversations. It doesn't really matter if it gets any of the words right, because you can press the play button and you can hear what it recorded, in your own voice, without any of the pauses when you listened to the other person . At the end of a business conversation, I can listen to my part to make sure I haven't forgotten about anything that I've committed to doing.

Instructions

1. Download the following files:
   1. Microsoft Speech Recognition Engine 1.0
      • Info

        Size	6 MB
        System Requirements	Pentium-120, Windows 95/98/NT [I'm running it on Win2000]
        CNet Description	This file consists of a speech recognition engine that provides speech input capabilities for Microsoft Agent. At this time Microsoft provides a single U.S. English speech recognition engine for use with Microsoft Agent.

      • Links
        • CNet's Page
        • A Direct Link to MS
   2. Speech SDK 5.1 for Windows
      • Info

        Size	68 MB
        MS Description	The Microsoft® Speech SDK 5.1 adds Automation support to the features of the previous version of the Speech SDK. You can now use the Win32 Speech API (SAPI) to develop speech applications with Microsoft Visual Basic® and other development systems that use Automation.

      • Link
        • MS's page
        • Direct Link
2. Run the downloaded files:

   During the installation of these two programs, I think it asks you to test your mic and speakers and do some setup-type work.

   1. I ran the SDK executable first, and I found that it was just a self-extractor. In doing this, just remember where you tell it to extract to! Then I went to that directory and ran the "setup.exe" file. This is what installed the SDK.
   2. Then I ran the ACTCNC.EXE file, which was more automatic. When this installs, watch the filenames that go flying by, so that you can see what to search for, because it hides out pretty well.
3. Open DictPad:
   • I'm installing on a Win2000 machine and it put it here:
     C:\Program Files\Microsoft Speech SDK 5.1\Bin\dictpad.exe

Tips From My Experience

• think of DictPad as Notepad with a voice menu and toolbar.
• To do more voice training, which I highly recommend, choose "voice training" from the voice menu.
• A very handy keyboard shortcut is Control+M, which will toggle on and off your microphone.
• Most punctuation marks and symbols found on the keyboard can be inserted by saying their name. With the "!", all you have to say is "exclamation." There is a way in which you can specify that it type out the word "period" or "exclamation", but I have not yet figured out how to do it reliably.
• You can also move the cursor around the document in DictPad. But you must be in "Command" mode. You can just say, "command" to enter it, from "Dictation" mode, or click the appropriate buttons in the menus or toolbar. Then you can use the move-around-the-document commands (which I should post soon, though you could probably figure out some of them).
• Speak at a normal face, unless you typically run words together as a habit. In that case, you may want to put a tiny gap between words. The key is that you speak as normal as possible, so that it will be able to recognize you as you speak in a normal way. If you train it speaking in a staccato, formal way, then to be able to use it, you'll have to speak that same way.
• While you're using it, don't try to say single, small words one at a time (I have tried it, and it's very frustrating). The program recognizes words best when they are used in context.
• Even though the voice menu has an option to add or delete specific words, don't use this for normal words. Let the voice training take care of that. However, you may want to train some proper nouns that you use often.
• You can also train symbol combinations like ;-) to appear for specific words, like "smiley face."
• Be careful as you're typing to correct any interpretational mistakes made by DictPad, because it usually interprets loud typing as words, and can mess up what you're typing.
• The "play" button will read the document to you in your own voice, making it easy to correct portions.
• The little gray square box below and to the right of the selection will give you a drop down list to choose words that the computer thinks that are alternatives to what you said.

Special Words

• Period - inserts a period and sets it to capitalize next word.
• Paragraph - inserts two line breaks and sets it to capitalize next word.
• New Line - inserts one line break and does not set it to capitalize next word.
• Dictation - switches to dictation mode.
• Command - switches to command mode.

---

Opinions

The following are opinions on the CNet page that may be of some use. Brackets, mine.

From: CNET > Downloads > Windows > Utilities > System Utilities > Microsoft Speech Recognition Engine > User Opinions

Install SDK first, then ACTCNC.exe. At this point you will have 3 more progs in you Start menu. One is Microsoft voice, for commanding you comp, the other is Microsoft dictation for dictating the text, and the last is Microsoft Speech Sdk for developers. [I did it this way but only ended up with the SDK and I couldn't find MS Voice anywhere.]

[The ACTCNC.EXE file also creates a "Speech" control panel but I don't know how exactly that affects anything .]

After downloading and installing, the pgm does do a hiding act but when you find it (c:\windows\speech), there are two exe's that need to be run depending on the type of app you want. I ran the "INSTMSIA.EXE" (another set-up pgm...) After running this, there is a program folder with various demos, voice training, and other cool stuff. [I don't believe I had to do this because I ran the setup.exe file instead of the one he mentions.]

Downloaded this file on www.download, installed it => I also didn't find where it was hiding (cf. neg reviews) I went to MS website, downloaded "SAPI4SDKSUITE" (39MB), installed it ==> installation asked for a cdrom which I didn't haven, so I cancelled. [I did not experience this, but I had SDK 5.1] However, two links were added in the start menu - programs. When I clicked it, I got an error (engine missing) I went back to the MS website and downloaded "speechsdk51" (69Mb) and installed it. The program still didn't work. So, I installed "microsoft speech recognition engine" again. If things went right, you can activate "microsoft voice" in the start-programs menu ...

you'll see that this product works great, if you follow their directions and get the required 69MB Speech SDK 5.1 which is located at http://www.microsoft.com/speech/ (win 98 only? see below for win 95 SDK 4.0 package) The 5.1 kit (probably also in the 4.0 kit) contains text-to-speech as well as voice recognition. The main voice recognition program seems to be the dictpad.exe file.[this is true.] It lets you see all its different guesses as it's "thinking" and is really neat. It (dictpad.exe) seems a lot less buggy than Via Voice and opens much quicker. You can train it for your voice in dictpad and train new words. Dictpad also has command and control capability mode, you can train words for it, but I don't see how you're supposed to make it perform windows actions for those words. The included text-to-speech program TTSApp.exe is better than ReadPlease in the sense that you have more control over the options. ReadPlease is using the same engine. The SDK 5.1 says it does not support win 95, so Win 95 users may want to use the 4.0 version instead. It's at http://www.microsoft.com/downloads/release.asp?ReleaseID=26299 and get it while you can for friends that have Win 95. I have Windows 98. I haven't figured out how to get it to type into any arbitrary application. [Me neither] One of the .exe tools in this kit let's you actually see the phoneme and word searches that are made as it tries to figure out what you said. As others have said, there is absolutely no basic documentation on how to use what's there, but there's a lot of documentation for showing programmers how to a lot of stuff.

As some people have said that the files are hidden; when you install this speech recognition software, the files are installed in c:\windows\speech\ and c:\Program Files\Common Files\Microsift Shared\SpeechEngines. [I just did a search for files named "speech" and it came up with all the applicable directories.]

---

Send me a message below if you find out any more helpful information that should be on this page.

For safety's sake, and because of bad people out there, I'm going to make it truncate anything past 3000 characters. If you need to send me something bigger, just contact me about it.

You are welcome to ask me for help, but please understand that I don't know what kind of response I'm going to get from this. So, understand that if I'm busy, and so often I am, I may not respond to your message at all...ever. I also may not have the foggiest idea about how to help you. ;-)

I put this page together in a hurry, and this should work, but It might not.

Your email:

Subject:

page created April 22, 2002