apple watch integration with voxcommando

[jitterjames]

Unfortunately, I can't really parse that block of text but I can tell you a bit about the xml that is returned by the simple web server when it receives a get request containing an action string.

You posted this xml, although I don't know if it was from an action that worked for you (as expected) or not:

Code: [Select]

<action>
  <actionString>VC.TellVox&&onkyo set volume 40</actionString>
  <Success>True</Success>
  <Result></Result>
</action>
This XML corresponds to what you should see in the history panel on the main VC window.  If the <Success> tag is "True" this corresponds to an action that is not shown with a red background in the history.  Based on this particular XML it means that your action was processed successfully.  That doesn't necessarily mean that anything will happen though because you would need to have a corresponding voice command with the same spelling.  I don't think VC.TellVox is case sensitive though.  Remember that this text has nothing to do with the command name, it must match the phrases and payloads defined for the command.  The VC.TellVox action has no way of knowing whether or not a matching voice command will be found, it only knows that it successfully emulated the text that you sent it.  You'll need to look at your history window to see what entries follow the action that the web server generates.

Different actions will return different xml.  Here are some examples:

Code: [Select]

http://127.0.0.1:8095/api/TTS.Speak&&Isn't+this+fun?returns

Code: [Select]

<action>
  <actionString>TTS.Speak&&Isn't this fun?</actionString>
  <Success>True</Success>
  <Result>speaking (asynchronous): Isn't this fun?</Result>
</action>In this case, this action returns a string for result.  This should correspond to what you see when you hover your mouse over this action in the history panel.  If you try it you'll see that it says:  info: speaking (asynchronous): Isn't this fun?
The word "info:" is to indicate that this result does not actually modify the {LastResult} variable.

Here is an example of a bad action string:

Code: [Select]

http://127.0.0.1:8095/api/TTS.Speaker&&Isn't+this+fun?this returns a value of "False" for "Success" because TTS.Speaker is not an valid action:

Code: [Select]

<action>
  <actionString>TTS.Speaker&&Isn't this fun?</actionString>
  <Success>False</Success>
  <Result>Error: unknown TTS action</Result>
</action>
And here is an example of an action that successfully returns a result that gets put into {LastResult}

Code: [Select]

http://127.0.0.1:8095/api/Gmail.GetCount

Code: [Select]

<action>
  <actionString>Gmail.GetCount</actionString>
  <Success>True</Success>
  <Result>9</Result>
</action>

[vulcanjedi]

Thanks for the insight James.
I tried some other commands that do work.
Interesting I've been having issues getting this long used command to work, audibly and tellvox too is.
All it is is a
tts.speak
and a
onkyo.setvoldecimal

I get that cant compile into cfg error although I saw a previous post that you said it wasn't critical.
Will continue to investigate, dont want to 'jack this thread.

[Gary_BBGames]

Hey Vulcan, having trouble understanding your post... is the Home Remote app working or not working?

[nime5ter]

Thanks for the insight James.
I tried some other commands that do work.
Interesting I've been having issues getting this long used command to work, audibly and tellvox too is.
All it is is a
tts.speak
and a
onkyo.setvoldecimal

I get that cant compile into cfg error although I saw a previous post that you said it wasn't critical.
Will continue to investigate, dont want to 'jack this thread.

If you decide you want help with this command, you can create a new thread and follow the standard practice for getting support: post the xml for the problem command and a log. As James indicated earlier, it's probably a problem with your command phrasing, not the macro.

[Haddood]

I got the app too ... Though no watch ... Testing with iPad and iPhone

[Gary_BBGames]

Hey Haddood, the watch app allows direct pass through of voice, the iOS app doesn't so you'll need to add commands. The iOS app and Apple Watch app use different voice recognition APIs, the watches being better. If you need any help setting the app up then let me know.

[nime5ter]

Thanks.

The way it will work is you specify the base URL in the settings, eg: http://192.168.0.30/whatever.xxx?

When you use voice recognition and press "done" it would append all that was recognised, URL encoded, to the base URL. (Functionality for this is Apple Watch only)

You can launch the app directly from the Apple Watch with a glance, going directly to voice recognition. You can see a video of the normal functionality here: http://youtu.be/poValYfohLg

Hi Gary,

Does Apple Watch's voice recognition API return just one result, no matter what?

I don't mean in terms of how your particular app is currently implementing it, but I am curious whether that API can, in theory, return a list of possible matches ranked according to likelihood.

[Gary_BBGames]

Hey nime5ter,

The Apple watch will return the text it shows on the screen before you press done, there are no other "possible matches" or confidence rating.

The value it returns is an array, but it's an array of just one string - the other entries in the array are for storing data of any emojis.

The voice recognition is very good, but with homophones like "weather" and "whether" there will be a problem.

If the VoxCommando support is well received and actually generates some income then I will add a feature to allow words to be swapped out on the fly, such as if the watch sees "knew" then use "new".

[nime5ter]

Thanks for the info, Gary. I was curious how it compared to Google's approach (which is as I described above -- as is Microsoft's).

For VC users, it's probably more efficient anyhow to edit all their commands in VoxCommando to contain alternate phrases in order to accommodate Apple's misinterpretations.

[jitterjames]

If the VoxCommando support is well received and actually generates some income then I will add a feature to allow words to be swapped out on the fly, such as if the watch sees "knew" then use "new".

I think there are quite a few VC users who long for a solution similar to VoxWav for the iPhone.  Probably not enough yet with watches to be a significant source of income, at least in the near future.

Maybe we should start a poll.  It would not help us to measure total volume but might give us an idea of % of people who use   iPhone || iPhone + Apple Watch || Android.

[Gary_BBGames]

I could add in the functionality for the Home Remote iOS app to do the same kind of thing that the Watch app does, although  due to the restrictions of iOS the app has to be open and in the foreground to listen in, limiting it's use compared to the VoxWav app (I checked it out online).

I could make it that if the voice recognition doesn't detect any inbuilt actions with the recognised name then I pass it through to the VoxCommando URL.

If you had a crappy iOS device you could leave it always on and listening, no data is sent online, it's all local recognition with the iOS app and, I believe, Apple Watch - I need to look in to what Apple's own APIs do.

[EDIT] THIS IS NOT POSSIBLE USING THE API I USE [/EDIT]

[vulcanjedi]

I'd be happy to try, I have a new iphone and an older iphone with the crappy broke button.
Still haven't managed to connect with sister to try her watch yet

I agree if this functionality was more akin to VoxWav on droid that'd be the best outcome and would think most would like the watch form factor for this.

If it helps Gary, you can spin up a virtualbox vm and free win10 tech preview to test with.

[Gary_BBGames]

Ah bugger, I've just been looking at the voice detection API docs for the API I use and you have to feed it a list of words to look for, this means I can't just let it loose listening to all and passing it on as it will try to match what it hears to the words in a sample dictionary.

[Kalle]

Hi Gary,
I have a "old" iPhone 4 and you can add me also as beta tester.

[Gary_BBGames]

Hey Kalle, not sure if you saw my later message, but using the voice recognition API that I currently use this isn;t possible