VoxCommando vs VoxCommandoSP?

[Daniel]

So I am not asking what makes them different. I understand that the SP version uses the Windows Speech Platform 11 and the regular one does not, but is there anyone that has tried both and found one to work better or recognize speech more accurately over the other? I was using SP when I first started but I switched to the regular one thinking it worked better for whatever reason. I think the problem might have been that it would stop listening to me after a while. However, now I thinking of trying it again because on the regular one I sometimes raise an eyebrow at what it thinks I said. Am I supposed to purge cache and rebuild grammar when switching between them? Which one do you use and why?

[jitterjames]

It is their differences that makes one better in some circumstances and another better in other circumstances.  If one were just better than the other then I would not have made both versions available.

Personally I prefer the regular non-SP version (for English) but there are situations where the SP version is better, or actually the only option.

The regular version can be trained to a specific persons's voice or you can use it with multiple voice profiles but you need to actively switch between them.  The SP version can't be trained but may be more flexible when dealing with different people speaking to it.  So if you are the only person using VC (and especially if you have an unusual accent) it probably makes sense to use the regular version and take a bit of time to train it.

The regular version supports dictation and SP does not so if you need dictation then your choice is pretty much made right there.

The SP version supports more languages, so for some people the SP version is the only choice.

SP also works better with Bluetooth microphones for some reason.  The regular versions really does not do well with most Bluetooth audio input.

You will sometimes get odd recognition that doesn't make sense based on what you said, but this is usually due to what is recorded and whether or not you've got a reasonable silence before your command.  With clean audio and a decent pause before you initiate your command you will usually get reasonable recognition.  You may also want to try turning your input levels down.  If you are getting a lot of improper recognition then it might help to create a new Windows speech voice profile and retrain it.  It is also worth doing this sometimes if you change to a new microphone.  Another way to improve the recognition accuracy can be to use a prefix if you are not already doing so.

[Daniel]

Is there anything I should do to train my voice or does it adapt to it on its own? Does purging the cache and telling it to rebuild missing grammar affect its progress? I am not 100% sure I know what I am doing when I tell it to do those things but I have started doing it when adding new commands or updating a payload because I don't know how to get it to recognize new things.

[jitterjames]

Is there anything I should do to train my voice or does it adapt to it on its own?

By default it does not adapt (learn) on its own.

I recommend using Windows built-in speech training.  This can conveniently be accessed from the VC main menu, under Options >> Windows Settings >> Speech recognition.

There are two other methods to train the profile directly from VoxCommando but under normal circumstances I would not use this.
- On the main menu under Options check the box for "Learn (Train your speech profile)".  This should work but be careful with it because the engine will learn "incorrect" recognitions as well.  That is why this is turned off by default.  In other words if VC thinks you said one command when in fact you said something else it will "learn" that this is how you say the command it thinks it heard, which would be bad.

- A more precise way to train VC is available from within the command editor.  Right click any command to train the speech engine that particular command phrasing.  In theory, this could be useful if you have a particular command that the engine does not recognize well.
See: https://voxcommando.com/mediawiki/index.php?title=Profile_Training_Tool

Usually I would just use the tools provided by Windows.  A single training session is usually enough. The other options within VC are more for special situations or if you have an unusual accent.

Does purging the cache and telling it to rebuild missing grammar affect its progress? I am not 100% sure I know what I am doing when I tell it to do those things but I have started doing it when adding new commands or updating a payload because I don't know how to get it to recognize new things.

Normally you do not need to purge the cache.  This is generally just a fallback if something goes wrong.  VC should identify which grammars need to be rebuilt and will just rebuild those ones.  The one time I would recommend using this is if you move your VC folder to a new folder or a new system.  At this point I can't even remember why this is so, but I am pretty sure it's a good idea in that case.  Also if you get errors with certain grammars and you want to try to force it to rebuild failed grammars you could use this, but normally it's not needed.  It's also not necessary to do a restart or a full restart under normal circumstances.  In some older videos you may see people doing this but the program is more intelligent about this stuff in later versions.