Deluxe Audio Strap + Voice Attack = 😡

I got my Deluxe Audio Strap some time ago and have really been enjoying it—with the big advantage being that I don't have to negotiate putting on the Vive, finding the headphones, putting them on, discovering they're on backwards and turning them around, etc. etc.—and that was after I'd bought a good wireless USB gaming set (SteelSeries 800) because, with wires, it was not uncommon that I'd try to strangle myself getting the wired ones off.


I'd been playing without Voice Attack lately, since before I got the Deluxe Audio Strap, because I've been livestreaming on PS4 and wanted to get used to doing things manually. So I hadn't had a chance to try Voice Attack with it.

I was about to embark on an RNGineer macguffin quest, though, and I'd be needing to operate in a bunch a different modes, with mining gear and collector limpets and SRV and so on—a gamestyle that Voice Attack with an HCS voicepack really excels at in increasing efficiency. So I started the game flatscreen with my wireless headphones to make sure everything was working properly since the last VA and HCS upgrades, then jumped into VR.


Please tell me this is not expected and normal behavior from the Deluxe Audio Strap's mic. I got recognition about 10% of the time or less. Checking the 100 lines of scrollback, just NINE are recognized commands, one of those was wrong, and most of them were "Yes" or "Confirmed" or "Engage".

For your amusement, I list what Voice Attack reported it heard below.

This was after I retrained (going through two extra training texts) my voice with Microsoft Speech recognition, using the Deluxe Audio Strap mic. I recorded some audio of myself talking with the Vive on. It doesn't sound great, but it sounds better than a telephone, with no drop-outs or static, so I have no idea why Voice Attack is having so much trouble.

I retried with the SteelSeries, and once again I got near-100% recognition, as I was used to.

Is this a known issue with this mic, or might I have a faulty module that affects Voice Attack somehow but is okay to human ears? (The obvious issue, that Voice Attack is listening to some other microphone, I ruled out—I'm getting level indicators on the Voice Attack window within a good range when speaking, and I verified that indicator jumps when I just tap the Vive lightly, so it isn't my webcam mic or something that's getting funneled errantly to Voice Attack.

This up-till-now great hardware has turned some up-till-now brilliant software into a perfectly unusable nightmare. Someone, please say you can help. o7

———

What I saw in scrollback:

✘ 'and cargo risk a'
✘ 'PG cargoes get to'
✘ 'one of its cars can'
✔ 'cargo scoop up' (Confidence 88)

✘ 'that it's a cruise to'
✘ 'so just a cruise'
✘ 'that's super cruise'
✘ 'soup for cruise'
✔ 'go to supercruise' (Confidence 60)


✘ 'Floyd assessed on event'
✘ 'live like assessed on'
✘ 'a light to us this to a often a'
✘ 'white cast a stone wall,'
✘ 'the latest assault up'
✔ 'flight assist off' (Confidence 67)
Except, that was supposed to be, 'flight assist on'...

✘ 'the latest installment'
✘ 'of a wily to assist one'
✘ 'talk with Clyde assessed'
✘ 'a wine glass list on'
✔ 'flight assist off' (Confidence 72)
That time, I did say 'flight assist off'.

✘ 'set a course on remedy'
✘ 'set a course our morality'
✘ 'set for a course we're up a'
✘ 'club record summer opening'
✘ 'set a course power up a'
✔ 'verity' (Confidence 95)
No, I didn't say "Verity"...

✘ 'set a course on wrap a'
✘ 'stagecoach Emerald bay'
✘ 'set course to repay'
✔ 'set course for merope' (Confidence 65)

✘ 'in excess the main route to'
✘ 'access to'
✘ 'lecture first Armenian rockets to'
✘ 'breakfast in loop'
✘ 'Lexus them enriched'
✘ 'legs as the star main ranch'
✔ 'next system in route' (Confidence 74)
 
Last edited:
Imho the mic on the vive is useless. Tried it with Voice Attack and it didn't recognize anything I was throwing at Voice Attack. I started to use an external mic via USB. Recognition is way better out of the box. Since I bought a Rift the other day I'm using now the built in mic on the Rift and I have to say this is an excellent mic.

In short: the built in mic on the Vive is pretty much useless. Use an external one.


Long days and pleasant nights
 
Yep, the mic on the Vive is rubbish, and notorious for being so - it's not the strap. When in the Vive i use a desk mic for VA. When in the Oculus I use the built in mic.
 
Last edited:
Ugh! Mildly frustrating.

Have you tried this?

Say the commands

Look at what VA thinks you have said

Add that to the phrase bank for that command

Problem solved? Or seriously reduced?

Edit
For toggle commands like FA or Cargo or landing gear, I just use FA or Cargo Scoop and let ED's voice do the interaction.
 
Last edited:
Deluxe Audio Strap has no mic. It's the vive's own mic.

Also I've heard that putting some mic foam (you know, the one used to eliminate noises) on the vive mic opening helps to reduce "punchiness" of the mic, that could help. Not that I noticed that punchiness at all while we're playing Arizona Sunshine coop with a friend, but maybe our language is less punchy, and the difference is noticeable by speech recognition algorithms.

Also my experience with Star Trek Bridge Crew voice recognition is also poor, although I thought that was a fault of my wooden English accent. May try it again with the foam covering the mic opening.
 
Also. EVERY time you add or change Mic for voice attack you need to train the speech engine in windows, and I recommend the full training. Not just the barebones "Peter dictates to his computer....."

And this need to be done in the manner you intend to use it.
For hmd's that means fully wearing the headset.

So you will need to do so in vr.
Using a virtual desktop environment is really the only solution.
I recommend Bigscreen, free for both oculus and steamvr

Most if not all voice recognition on a computer uses this same piece of windows, so it would most likely help with games and software that uses it.
 
Last edited:
Also. EVERY time you add or change Mic for voice attack you need to train the speech engine in windows, and I recommend the full training. Not just the barebones "Peter dictates to his computer....."

Wait just a minute... It erases the model every time a new device is plugged in? So for example when I plug my headphones with mic to play some Overwatch it actually erases the model I used for the Vive? O_O
 
Wait just a minute... It erases the model every time a new device is plugged in? So for example when I plug my headphones with mic to play some Overwatch it actually erases the model I used for the Vive? O_O

It wouldn't surprise me if it didn't, but I'd hope Widows would keep a profile for each mic. I only changed my mic once, when switching from a headset to the Rift, and I set up the mic for speech recognition and trained it a few times. So don't just train it, but also set it up for proper levels - all part of the speech recognition page in the control panel.
 
Wait just a minute... It erases the model every time a new device is plugged in? So for example when I plug my headphones with mic to play some Overwatch it actually erases the model I used for the Vive? O_O

No poorly wording on my part maybe.
But for each microphone you will have to run training and that should be ok until drivers are removed or re-installed/updated.

If you used a headset with microphone before. And just now start using the vive Mic then it would need training.
But you should be able to switch between the two without much hassle after they have been sufficiently trained.

And like I mentioned.
Position of the microphone is hugely important.
If you intend to use it while in vr it also must be calibrated from within vr.
 
Last edited:
If you intend to use it while in vr it also must be calibrated from within vr.

You're making an awful lot of rather ungenerous assumptions that, when I wrote:

This was after I retrained (going through two extra training texts) my voice with Microsoft Speech recognition, using the Deluxe Audio Strap mic.

because I mentioned training my voice "using the Deluxe Audio Strap mic", when the audio strap doesn't have a mic, that I must have done something completely different, and was badly confused about my own actions or lying about them.

The charitable explanation would be that I did retraining in VR, and had just falsely assumed that the thing called a "Deluxe Audio Strap" would have an "audio" component like, say, a "microphone". The charitable explanation also happens to be the true one—I used Bigscreen to magnify my desktop and put the training window in the center (as an ultrawide desktop is very bad when projected into VR no matter what you do). I did two sets of extra training—I'm not sure how many counts as "full". I'm sure I didn't do more previously.

(I'd never previously used or trained the Vive mic, because, as I wrote, I used a high-quality wireless USB headset with built-in boom mic.)

I'll give the foam a try, and if that doesn't work will give a good-quality lapel mic a try (though I can't plug it into the Vive without a Y-splitter that will have to dangle around my head or be taped down, and it won't reach to the computer chassis so I'll have to do something like that—but you can't have everything, and if I dedicate a Y-splitter I won't have to pop the port cover door off the Vive every time I need to use the mic for something different).

andysonofbob, if you look at those actual examples in my OP, I can't imagine entering them all in—there are 800+ speakable phrases in the HCS voice pack and as I wrote above, it got 5–6 incorrect guesses when I repeated each of just a few of them. And the wrong guesses are so random that I don't think it'll have much effect—it didn't hear me say the wrong same thing twice. As for just saying "Flight Assist" or "Cargo Scoop", rather than adding "on/off", "up/down", "open/close", I'll try that—my assumption was that more phonetic data (i.e., longer phrases) would give it a better chance of identification (I was assuming what we in computational linguistics call a "primed" or "vocabulary" model), but that could be backwards with how Voice Attack works (it could be speech-to-text or speech-to-phoneme followed by a "best-recognition" model), in which case shorter commands will be recognized better. (It thought I was saying "Punch it" when I was saying "Engage", and "Go", when I said "Now"—luckily those all work when you successfully get HCS packs to recognize an "on my mark" command—and this was what made me think it was using the vocabulary model.)
 
Last edited:
Can someone recommend me a reasonably priced desk mic?

No—I have no experience with desk mics below $100 (the Audio-Technica AT2020 is what I use), but the Røde SmartLav+ (Amazon link here) is a great general-purpose lapel mic. I haven't tried it in all configurations with the Vive yet, but if you have the Deluxe Audo strap, you may need a Y-splitter and an adapter so that it will work with the Deluxe Audio Strap's headphones—looking at the ports and bands on the plugs, I think you may need both, but I can't try right now because the hubbie is gaming—my schedule is more flexible than his, so I let him have the PC on weekends. :)

...I think you may need both...

And should add, "but both will definitely work", if you don't want to buy just the mic and find it isn't sufficient and then have to buy an adapter too. (Update: see below; this doesn't work, unfortunately, not for the Vive.) I use all three for my PS4 streams and they work great.
 
Last edited:
And should add, "but both will definitely work", if you don't want to buy just the mic and find it isn't sufficient and then have to buy an adapter too. I use all three for my PS4 streams and they work great.

NO, I was wrong. After a chance to actually try it, it turns out the Vive's 3.5mm jack is just a headphone jack. It will not take microphone input, even with the appropriate adapters. [woah]

So... the only option here is a mic that will connect to the computer and stretch sufficiently for wherever you want to use the Vive. I guess that's sufficient for Elite: Dangerous; not so much if I wanted to use it for other games. Damn.
 
And I just had an opportunity to test it, and the Røde SmartLav+ (with the adapter, plugged into the motherboard mic input port) works great with Voice Attack—I just had to do the first, shorter, training, not the additional two, and it responded with near-100% accuracy once again. If I get an extension cord, it should be fine, if not ideal to have yet another cable to get caught up in..

(It does seem quite a bit slower than I remembered—I distinctly remember using "seventy-five percent" as my way to begin approach when I saw ETA reach 0:09 or so, and it's now taking too long and variable a time for that to really work... perhaps lag from the HCS plugin, which I didn't use before? I don't know.)
 
This is why personally I plan on using my sennheisers even with the audio strap. The strap is probably worth it just for comfort, but I like my big expensive headphones thanks...
 
This is why personally I plan on using my sennheisers even with the audio strap. The strap is probably worth it just for comfort, but I like my big expensive headphones thanks...

If your play style is put on the Vive, play, turn off VR, take off the Vive, that's probably sufficient. If I could consistently remember to up my UI size dramatically and/or reduce my screen's effective resolution before starting VR it would probably work for me, too, but too frequently I need to make adjustments in Windows, and you can't run a program that displays the Windows desktop in a useful format (like BigScreen) concurrently with Elite: Dangerous. You can run OVRDrop (highly recommended) for EDDB, Inara, etc., as virtual screen(s) in the virtual space of your cockpit or SRV, but that only works for things you know before play you'll need. If I have to adjust Voice Attack or system settings during play, I have to take the HMD off, and the Deluxe Audio Strap makes that much easier, but only if you use the on-strap headphones. Remove them, and you're still struggling with getting the headphones off first. (Also, the Deluxe Audio Strap's fit-tightening is going to get in the way of many headsets' bands.)

[font=AvenirNext-Medium,Roboto]Protip: With OVRDrop, tie a window to each of your two motion controllers, and in OVRDrop settings, make the "back" of the overlay "None". The motion controllers can be used as a "virtual tablet" for reference sheets and other things that you just need to read and scroll; when you don't need them, flip the motion controller around and the virtual screen disappears until you flip it again.[/font]​
 
Last edited:
Back
Top Bottom