Vista Speak Easy

I'm listening...

I'm presently marking loads of software submissions from our second year course. This involves looking at some designs, making some comments, coming up with a mark and them moving on to the next one.

A hundred times.

I thought that rather than type everything I'd try speaking it instead. So, for the first time, I hooked a microphone up to the computer and fired up Vista's speech recognition.

I didn't have particularly high hopes. The last time I saw it being used was when a hapless presenter tried to show off a beta version of windows at a talk they were giving. The results were highly amusing, and probably the product of beta code and a dodgy microphone setup. However, they did serve to put me off trying to use the system (although it is very funning when the presenter says "delete sentence" and the system dutifully puts "delete sentence" into the text).

The training session is interesting. You go through learning how to control the program whilst at the same time the system is learning how you speak. This means that you can say profoundly wrong things as you train it, and it still seems to work (although this will come back and haunt you later when it uses the trained data to try and make sense of your real speech).

It took a while to complete all the training tasks, but there are a large number of options and you really could use the speech interface to control pretty much all of the machine. The text correction stuff is very clever, and makes it easy to correct particular errors.

Then it was time to use the system in anger. And it worked pretty well. I could just dictate comments and they are decoded and fed them into the window where the cursor happens to be. I didn't find any particular need to speak more slowly, the system actually seems to work better if you throw a whole load of text at it rather than single words - probably because it uses a lot of extra context information from the text to decode the sounds. I knew I was on to something when I started using the voice input to write and send an email. The only problem is that you have to compose the whole sentence in your head before saying it, and this is not usually how I write.

Having said all this, I'm definitely going to get a proper microphone and start using the speech input as part of the way I work. If you've never tried it I'd recommend it, I'm not sure which versions of Vista it is supplied with (I'm using Ultimate - which seems to have everything) but if it is there it is definitely worth a go.

Comments

[Richard]

Hello Rob.

Um, can you tell me how you would make a pixel fire out of a piece of graphics? A graphics that you can move around I mean.

[Richard]

I am also using Visual studio 2005, visual basic express edition. When I try and port some sort of small program for my pocket pc I'll do it at my old college when possible.

[Rob]

Hmmm. Making a pixel "fire" out of a piece of graphics is not actually what you do. Effectively you need two sprites, one is the thing that is firing (the gun if you will) and the other is the bullet which is being fired. When the bulet is fired you simply make it start at the appropriate position on the gun (perhaps the end of the barrel) and then move away from the gun in the appropriate direction. To the player it looks as if one thing has fired another, but actually we are just managing two sprites. I do this kind of thing in the Smartphone starlight program, where the rocket can fire missiles.

[Richard]

Would you be able to give me a small example in Visual basic?

If I have this piece of gfx,

e.graphics(pens.white, 10, 10, 100, 100)

e.graphics(brushes.white, 10, 10, 100, 100)

These pieces of code make that, so if you can move that around using the buttons etc

Um, firing a pixel out of that shape erm...? I really have no clue, and kind of find that tut on msdn a little nuts and bolts....

[Rob]

This is a funny place to put a discussion on game creation. But never mind. I think the key is understanding what the e.graphics(brushes.white, 10, 10, 100, 100) does. I think you will find that two of the numbers (perhaps the 10 values) will be the draw position and two the size. If you arrange for the draw operation to be repeated then you can move the draw position so that the object appears to move. That is how my Smartphone with Cheese stuff works.