Coffeehouse Thread

16 posts

Forum Read Only

This forum has been made read only by the site admins. No new threads or comments can be added.

Vista speech recognition video, It actually works

Back to Forum: Coffeehouse
  • User profile image
    longzheng

    Contrary to what MSNBC criticize as a ‘wreck’, the speech recognition technology is well developed and highly usable. I got my hands on the July CTP build (5427) of Windows Vista and gave it a try, and I recorded what I found. I used the internal microphone array in my laptop, so the sound quality is not optimal but Vista handled it well.

    Watch the video/screencast

    Enjoy Smiley

  • User profile image
    JohnAskew

    I love speech technology.
    What I'm not crazy about is the new SAPI.dll is not managed code.

    My own experiment with SAPI.dll 5.1 was very similar to the infamous recent demo.

  • User profile image
    BryanF

    Well done. Thanks. Smiley

    Larry Osterman explains what went wrong at the FAM here.

  • User profile image
    longzheng
  • User profile image
    Human​Compiler

    Nice video!  It's looking really good.  Seems kind of pointless to have to say "OK" after that dialog of possible matches comes up, but saying something on a web page and then it giving you numbers to say for exactly which one you want is hott!

    I love the "stop listening" or whatever it was at the end.  It feels so Star Trek.

    "Computer, locate Lutenent Warf."

    Can I make it take single commands when it hears me say the word "computer"?  I'm geekin' out here!  Tongue Out

  • User profile image
    eddwo

    As a test I managed to play a complete round of the Purble Place pattern matching game using speech recognition.

    It took a lot longer than using the mouse, but it worked pretty well.

    Basically lots of, "Press left", "Press up", "Press enter", "Press down three times", "Press enter"

    I like the way the speech recognition training was combined with the tutorial. Saves a lot of time over just reading a section from a book and then having to learn what features are available in the software.

    Can anyone manage the Purble Place cake making game on hard level with mulitple cakes in flight? I can get the first couple of cakes made, but when it stops you shuffling them back to the start it gets difficult really quickly. Wink

  • User profile image
    erik_

    THIS ROCKS!

    Didn't see this live yet, this is way better than the speech in XP.

    Does this works for every program or does the program needs to be speech supported to get the same effect as on solitaire (the 'show numbers' command)?

  • User profile image
    Steve411

    "I'm speechless." Wink

  • User profile image
    longzheng

    erik_ wrote:
    

    Does this works for every program or does the program needs to be speech supported to get the same effect as on solitaire (the 'show numbers' command)?



    Works for every application coded correctly.

  • User profile image
    Sven Groot

    HumanCompiler wrote:
    Nice video!  It's looking really good.  Seems kind of pointless to have to say "OK" after that dialog of possible matches comes up,

    I suppose it's so you get a chance to correct it if it picks the wrong number.

    HumanCompiler wrote:
    I love the "stop listening" or whatever it was at the end.  It feels so Star Trek.

    "Computer, locate Lutenent Warf."

    I think you meant Worf. Tongue Out </nitpicky>

    Coincidentally, the first speech recognition system I ever played with came with a Star Trek product. The Star Trek Omnipedia could be voice controlled using software developed by Dragon. It was absolutely terrible. If there's anything the Vista speech recognition is unbelievably good at, it's recognizing single letters (and in the rare occasion it gets it wrong you can use stuff such as "c as in cat" to make it clear). The Omnipedia always got it wrong. Whichever index entry you wanted to jump to by saying its letter, it was simply always wrong. It would also recognize half of what you say as "previous entry". You could sneeze in the microphone and it'd think you said "previous entry".

    It also didn't work under Windows 95, only Windows 3.1. This could be worked around by copying dragon.dll from a slightly newer Simon & Schuster product (Star Trek Klingon, which used speech recognition to grade how well you pronounced Klingon words; useless but fun) which did work. It didn't get any more accurate for it. Smiley

    HumanCompiler wrote:
    Can I make it take single commands when it hears me say the word "computer"?  I'm geekin' out here! 

    On a serious note, it would be useful if there'd be some way to indicate that your next spoken words are commands and not dictation of vice versa. It's very frustating to say "bold" in Word and see it write "bold" in the document. I wonder how it determines whether you're commanding or dictating anyway. Prefixing commands with "Computer" would be good enough for me as a distinction in most cases. Smiley

  • User profile image
    apoc

    as a last resort you can use show numbers in all apps

    btw, windows media player 11 isn't 100% speech ready, the playlist is almost impossible to control with "say what you see".

  • User profile image
    Yggdrasil

    Sven Groot wrote:
    
    On a serious note, it would be useful if there'd be some way to indicate that your next spoken words are commands and not dictation of vice versa. It's very frustating to say "bold" in Word and see it write "bold" in the document. I wonder how it determines whether you're commanding or dictating anyway. Prefixing commands with "Computer" would be good enough for me as a distinction in most cases.


    And if you wanted to dictate the word "computer"?
    For VB folk, saying "computer computer" would probably work. If you're more of a C-language fan, say "backslash computer".

    Maybe just emit some sort of high-pitched whine to specify the next word is a command.
    "I must speak with the dolphins now. Wheee! Wheee! Wheee!"

  • User profile image
    longzheng
  • User profile image
    jmacdonagh

    longzheng wrote:


    That's funny. But really, what else is it going to do?

  • User profile image
    cheong

    jmacdonagh wrote:
    

    That's funny. But really, what else is it going to do?

    I suppose it should name tabular cells A1, C2, etc... but then there's often no obvious hint that those buttons are "in table".

    So I think it has done it's best job.

    Recent Achievement unlocked: Code Avenger Tier 4/6: You see dead program. A lot!
    Last modified
  • User profile image
    Human​Compiler

    Sven Groot wrote:
    
    HumanCompiler wrote: I love the "stop listening" or whatever it was at the end.  It feels so Star Trek.

    "Computer, locate Lutenent Warf."

    I think you meant Worf. </nitpicky>



    Ack, I'm so embarassed!  Embarassed  I'm such a DS9 fan.  What have I done?  The horror!  [C]

Conversation locked

This conversation has been locked by the site admins. No new comments can be made.