Keyboard Shortcuts
Likes
Search
TSR-310 voice recognition
开云体育Hi, i?m just experimenting with the built-in voice recognition feature of the TSR-310. I?m talking about using the “Voice Control” device extender for the TSR-310. Does anybody know what service is behind that? ? Does anybody have more insight into that? ? Cheers, Thorsten |
|
It's basically the "OK Google" cloud service.? The trick is that it's in free-form/dictation mode, so if you have a scene called "wakeup" it may come through as "wakeup", "wake up", or "makeup"? :)? It doesn't know what the valid commands are, so you could get anything back. On Thu, Apr 29, 2021 at 5:36 AM Thorsten K?hler <thorsten.koehler@...> wrote:
|
开云体育Ah – thanks for the information. ? Cheers, Thorsten ? Von: [email protected] <[email protected]> Im Auftrag von Jeremy Weatherford ? It's basically the "OK Google" cloud service.? The trick is that it's in free-form/dictation mode, so if you have a scene called "wakeup" it may come through as "wakeup", "wake up", or "makeup"? :)? It doesn't know what the valid commands are, so you could get anything back. ? On Thu, Apr 29, 2021 at 5:36 AM Thorsten K?hler <thorsten.koehler@...> wrote:
|
|
It does require the voice services setup on the processor and last I messed with it you had to have an Alexa account linked as you would with regular Alexa integration. I had many issues with it actually working correctly though. I could speak the exact same command to the remote as I would to the Alexa device, and the Alexa commands worked 100% while the remote commands would almost always fail. I opened a case with Support and they were able to reproduce, and it just never went anywhere. Dont know if they fixed any of it but we gave up on the voice remote finctionality and instead direct clients to just use Alexa devices as its a much more reliable experience.?
|
开云体育I am using the amazon service with mine.? It works not well enough to deploy at a client site, but it is cool when it works. ? From: <[email protected]> on behalf of Jeremy Weatherford <jweather@...> ? It's basically the "OK Google" cloud service.? The trick is that it's in free-form/dictation mode, so if you have a scene called "wakeup" it may come through as "wakeup", "wake up", or "makeup"? :)? It doesn't know what the valid commands are, so you could get anything back. ? On Thu, Apr 29, 2021 at 5:36 AM Thorsten K?hler <thorsten.koehler@...> wrote:
|
开云体育Can?t confirm this. I tried with an empty program with just the TSR-310 and then added the “Voice Control” device extender. Compiled, uploaded and then voice recognition worked out-of-the-box without any service setup. ? ? ? Von: [email protected] <[email protected]> Im Auftrag von johnh via groups.io ? It does require the voice services setup on the processor and last I messed with it you had to have an Alexa account linked as you would with regular Alexa integration. I had many issues with it actually working correctly though. I could speak the exact same command to the remote as I would to the Alexa device, and the Alexa commands worked 100% while the remote commands would almost always fail. I opened a case with Support and they were able to reproduce, and it just never went anywhere. Dont know if they fixed any of it but we gave up on the voice remote finctionality and instead direct clients to just use Alexa devices as its a much more reliable experience.? |
|
After begrudgingly having to do a few of these, here's my take on how it works. There are two TSR-310 Device Extenders for voice processing...
toggle quoted message
Show quoted text
The first, "Voice Control", is basically the raw speech-to-text data from the microphone. It can be used alone, without relying on "the cloud", but then programming must account for every possible interpretation, as Jeremy described below. The second, "Dialog Client", takes the raw speech text and sends it to EITHER Amazon or Google's servers for processing. It basically figures out not only what was said, but what was intended, such as "too", "to", or "two" - then either acts on it using the Alexa or Google Assistant cloud service. The remote essentially becomes another Amazon Echo or Google Hub device, and the processor and/or remote must be registered for the appropriate service. I say "and/or" because it's been a year or two since I last did one, and can't remember. Although I'm still not a fan, the few I did worked reasonably well for what was required. The key is setting client expecations :-) One cool by-product is the processed commands (what Amazon or Google determines was said) are EXPOSED in the Simpl program as a serial string, allowing you to do things with a simple SIO that couldn't be done with Alexa alone. --- On Thursday, April 29, 2021 at 7:52 AM, Jeremy Weatherford wrote: |
|
Ah, thanks for clarification, Tony.
toggle quoted message
Show quoted text
I was indeed only experimenting with the " raw speech-to-text data from the microphone", hence the "Voice Control" extender. Indeed, you need to do the parsing and processing yourself, but for just some short one-word commands, it seemed to be pretty reliable, even with German language (after switching to "de"). -----Ursprüngliche Nachricht----- Von: [email protected] <[email protected]> Im Auftrag von Tony Golden Gesendet: Donnerstag, 29. April 2021 16:45 An: [email protected] Betreff: Re: [crestron] TSR-310 voice recognition After begrudgingly having to do a few of these, here's my take on how it works. There are two TSR-310 Device Extenders for voice processing... The first, "Voice Control", is basically the raw speech-to-text data from the microphone. It can be used alone, without relying on "the cloud", but then programming must account for every possible interpretation, as Jeremy described below. The second, "Dialog Client", takes the raw speech text and sends it to EITHER Amazon or Google's servers for processing. It basically figures out not only what was said, but what was intended, such as "too", "to", or "two" - then either acts on it using the Alexa or Google Assistant cloud service. The remote essentially becomes another Amazon Echo or Google Hub device, and the processor and/or remote must be registered for the appropriate service. I say "and/or" because it's been a year or two since I last did one, and can't remember. Although I'm still not a fan, the few I did worked reasonably well for what was required. The key is setting client expecations :-) One cool by-product is the processed commands (what Amazon or Google determines was said) are EXPOSED in the Simpl program as a serial string, allowing you to do things with a simple SIO that couldn't be done with Alexa alone. --- On Thursday, April 29, 2021 at 7:52 AM, Jeremy Weatherford wrote: |
|
Oh, let me add:
toggle quoted message
Show quoted text
As described, even with just the "Voice control" extender, the remote seems to communicate with a Google server according to my firewall logs. -----Ursprüngliche Nachricht----- Von: [email protected] <[email protected]> Im Auftrag von Tony Golden Gesendet: Donnerstag, 29. April 2021 16:45 An: [email protected] Betreff: Re: [crestron] TSR-310 voice recognition After begrudgingly having to do a few of these, here's my take on how it works. There are two TSR-310 Device Extenders for voice processing... The first, "Voice Control", is basically the raw speech-to-text data from the microphone. It can be used alone, without relying on "the cloud", but then programming must account for every possible interpretation, as Jeremy described below. The second, "Dialog Client", takes the raw speech text and sends it to EITHER Amazon or Google's servers for processing. It basically figures out not only what was said, but what was intended, such as "too", "to", or "two" - then either acts on it using the Alexa or Google Assistant cloud service. The remote essentially becomes another Amazon Echo or Google Hub device, and the processor and/or remote must be registered for the appropriate service. I say "and/or" because it's been a year or two since I last did one, and can't remember. Although I'm still not a fan, the few I did worked reasonably well for what was required. The key is setting client expecations :-) One cool by-product is the processed commands (what Amazon or Google determines was said) are EXPOSED in the Simpl program as a serial string, allowing you to do things with a simple SIO that couldn't be done with Alexa alone. --- On Thursday, April 29, 2021 at 7:52 AM, Jeremy Weatherford wrote: |
|
I've played around with the TSR-310 voice recognition and actually have been pretty impressed with it.? I have offered it to a few clients but for the most part, clients have not been that interested it. They just want a remote to control their AV system and the TSR-310 does that well with or without voice. So I haven't done any in field installations yet.
I found it is not necessary to link to a google or amazon account.? Just registering with is enough.? Other report better voice recognition with a link Amazon/Google account.? But for most things I have used it for, I found that is not necessary. I found that when you pair the TSR-310 with the processor, only the first one will pair.? When you try to pair the others, they will fail.? But no big deal because after you pair the first, all the other will work too even though they say they failed the pairing.? Like Tony mention, the big advantage is the serial string youvcan send to a SIO symbol and create any type of custom voice prompt you want. Steve |
开云体育I did, and i?m very interested – however, not available in Germany yet… ? ? ? Von: [email protected] <[email protected]> Im Auftrag von xsdesignstudios@... ? The prompt works fairly well, but it is way easier to integrate with Josh.ai since you can assume multiple phrases for something like wakeup, makeup, etc on the serial string.? If you haven't heard of Josh.ai yet.? Go look now. |
开云体育Hm, i?m just wondering why it wasn?t required for me to register or pair anything – it just worked with adding the voice control extension to the program, nothing else. And I also found the recognition rate pretty good and the fact that you just can grab the output string and do with it whatever you want. ? ? Von: [email protected] <[email protected]> Im Auftrag von Steve Kuehn ? I've played around with the TSR-310 voice recognition and actually have been pretty impressed with it.? I have offered it to a few clients but for the most part, clients have not been that interested it. They just want a remote to control their AV system and the TSR-310 does that well with or without voice. So I haven't done any in field installations yet. |
|
Well that is interesting.? To be honest, I just assumed you have to register it with fusion to get it to work.? Are you complaining that it works with out registering? If would be nice if all our problems were that severe.
toggle quoted message
Show quoted text
Thorsten K?hler |
开云体育I?m not complaining, just wondering… ? ? Von: [email protected] <[email protected]> Im Auftrag von Steve Kuehn ? Well that is interesting.? To be honest, I just assumed you have to register it with fusion to get it to work.? Are you complaining that it works with out registering? If would be nice if all our problems were that severe.
|
开云体育Thanks for confirmation – no I can sleep well again ? ? ? ? Von: [email protected] <[email protected]> Im Auftrag von dblpnt ? No need to wonder, the Voice Recognition Extender is using Googles Speech to Text engine as every other TSW, you only need to pair with Crestrons voice portals if you want to integrate with Amazon Echo/Google Assistant. |
|
If I understood i correctly the Googles Speech to Text engine is something build in the Android operating system.
So it's internal the decive and TSWs and TSR are Android devices. dblpnt, mainly wrote this so :) believe mostly it was understood. I was trying it with TSR-302 remotes and with TSW-xx52 panels, but it was not useable in my opinion I was hoping at release. Not shure if it changed in any way. |
开云体育Hi Stefan, ? I was hoping the Speech to Text is working solely locally on the device, but as i mentioned, i noticed in my firewall logs that the TSR-310 is contacting Google-servers. So i?m not sure about that. ? I did some test with a TSR-302 some time ago with mixed results, but IMHO that was more because of the hardware (functionality was unreliable), not the software. That seems to be better with the TSR-310. ? Cheers, Thorsten ? Von: [email protected] <[email protected]> Im Auftrag von stefan.mehre ? If I understood i correctly the Googles Speech to Text engine is something build in the Android operating system. |
Dave H - Crestron True Blue Tech Support Engineer Since 2014