Performing Actions Faster

[Stargazer]

I've been using VoiceAttack for gaming, but I've been finding it somewhat slow and clunky. Basically, when I give a command, it seems to take roughly 4-5 seconds: it takes ~2 seconds to recognize what I said as a command, then ~2 seconds before I can issue a new command (if I try to say something too soon, it seems like it gets mixed up with the first command).  All in all, in games this often ends up with me standing still for 5 seconds and dying while waiting for the command to go off.


Is there a way to make things "faster"?  Both in terms of how quickly the program recognizes my commands, and how quickly I can say things back-to-back.  Is it possible to set it up so that I can say multiple commands almost instantly back-to-back, or will I always need the 1-2s delay between them?

Right now I have the "Recognized speech delay" set to 0.

Also, with respect to training: is there a way to train my specific commands, rather than doing the general training through windows?  I feel like it would help more if I trained it on what I'm actually going to be saying, rather than their default prompts.

[Pfeil]

There is an experimental feature in beta v1.7.3.4 where you can set commands to essentially be recognized quicker, without having the speech recognition engine wait for you to finish speaking.
If you have that version, use the "Recognition" dropdown in the "Add/Edit a Command" window to set this option.

The options are now, 'Normal' (speech events processed fully as they've always been), 'Continuous Speech', which represents what 'Attempt recognition within continuous speech' was previously, and(new) 'Restricted Continuous Speech', which attempts to recognize commands as part of continuous speech but ONLY if the spoken phrase is at the start of the speech event.

[Gary]

If a recognized command is taking that much time to execute, you're either looking at a speech engine problem or a noise problem.  You should be getting sub-second responses even on a slower system.

You'll want to take a look at these threads:

https://forum.voiceattack.com/smf/index.php?topic=64.0 (ignore the title)
https://forum.voiceattack.com/smf/index.php?topic=63.0
and
https://forum.voiceattack.com/smf/index.php?topic=1635.0

I would definitely try the top thread first, and train your speech engine.

Also, there are alternate speech engines available:  http://www.voiceattack.com/alternate

Hope some of that helps!

[Stargazer]

There is an experimental feature in beta v1.7.3.4 where you can set commands to essentially be recognized quicker, without having the speech recognition engine wait for you to finish speaking.
If you have that version, use the "Recognition" dropdown in the "Add/Edit a Command" window to set this option.

The options are now, 'Normal' (speech events processed fully as they've always been), 'Continuous Speech', which represents what 'Attempt recognition within continuous speech' was previously, and(new) 'Restricted Continuous Speech', which attempts to recognize commands as part of continuous speech but ONLY if the spoken phrase is at the start of the speech event.

Where exactly is this option?  I have the beta, but I can't seem to find that dropdown anywhere.  In my general settings, it is definitely not on the "recognition" tab, and I don't see a "Recognition" dropdown anywhere when I go to add or edit a command.

[Pfeil]

See attached.

If you don't have that option, make sure you're on beta v1.7.3.4 or v1.7.3.5, as earlier versions don't have it in that form.

[Stargazer]

See attached.

If you don't have that option, make sure you're on beta v1.7.3.4 or v1.7.3.5, as earlier versions don't have it in that form.

Thanks a bunch - the button is so small and nondescript, that I apparently just kept glossing right over it. 

[Gangrel]

ONe thing I have noticed is that depending on the "window" that you are sending commands to, that Sending commands to the "Active window" (via the target dropdown box) can result in stuff happening faster than if the Target dropdown box is set to a specific application...

[Gary]

If you supply a value in the, 'target' box, VA will keep searching for your target like this (in this order):

Window title
Process name
Class name

Each one of these will require finding the item, setting the item to the foreground and then performing the actions in the command.

If you want VA to give up after looking for a window title (and not search for processes or class names), you can run VA with the '-noclasses' and '-noprocs'  command line parameters (as of beta 1.7.3.1).  We're only talking milliseconds here, though.  So, although adding these to the mix to make things fasterr (as well as using the continuous/restricted continuous option), we're not talking about fixing the huge 2 second delay - that is going to be some kind of noise (line or environmental) or speech engine component issue.  Just want to be clear about that.