Advantages of Built in Voice Recognition vs Alexa or google home

[AshaiRey]

Googling some found this...maybe it will work for you.

Cortana is available in these regions for these languages:

Thx Pete.
For HS2 I developed my own speech recognition application based on this however the recognition of Dutch is not that good. Atleast far to low to have enough WAF points to proceed.

What I hate is to rely on cloud services all the time.
I am planning to develop a HS3 app that only will trigger on the attention phrase and then, and only then switch to an online server that is better equipped for Dutch or any other language.

But, so much other things to do and just one life.... sigh

[Pete]

Yeah...looking today like all or most of the old Speech companies have consolidated and now doing everything in the cloud. That said though computers are much faster today than yesteryear and can do the processing without a cloud connection.

Wondering what ABNAMRO is doing these days. I was involved in some speech stuff back when they were in Chicago. The new stuff (speech and VR) most likely doesn't utilize the cloud today.

Relating to touch screen kiosks stuff it was never any chatty kathy stuff but rather biometics that was pursued at the time and it worked great.

I like the TTS but really have never utilized VR much. I like what the Amazon Echo does.

That said though I prefer a quiet touch screen interaction (mostly utilized for status these days if at all) for my automation over a chatty kathy always listening device on my network.

It is fun to watch people talk to and listen and do things that their smart phones tell them to do. Imagine what the future holds.

Must be my age.

[Rupp]

To stay on topic, there is no comparison between the Microsoft voice recognition (which is what HS3 uses) vs Alexa/Google. The Alexa/Google VR is light years ahead of Microsofts VR to the point where the later will most like disappear in the next few years. Do yourself a favor and get a Dot and give it a go. Instant recognition with zero training.

[Pete]

@Rupp...on a bit of a tangent...dutch VR / TTS was brought up and the lack of Alexa VR/TTS devices.

That said when installing Windows 10 Redstone initial TTS / VR was coming from the OS before it connected to the internet. No training easy peasey stuff.

I do not believe that Microsoft SAPI will be NOT be going away soon and probably will be integrated to use or not to use the cloud.

But this is me guessing right now...

I am seeing that the OP wants to use TTS / VR without the cloud. That said new Microsoft SAPI today is much better than yesteryears' SAPI.

Any thoughts as to advantages and disadvantages to both are greatly appreciated!

The list of advantages and disadvantages have been presented whether using cloud based TTS / VR or Microsoft based SAPI TTS / VR since the inception of initially mobile VR applications, then the first Alexa and followed by GV.

Text to speech was utilized much with Homeseer 1 and Homeseer 2 by many users. Voice recognition was not really used much and there were a few Homeseer users that did utilize it and yes it was frustrating to teach Homeseer using Microsoft SAPI VR.

Then came Alex and Google VR and it was easy and it works but it comes with the acceptance of enabling a microphone to the cloud which mostly many folks do not care about.

My point here though is not to abandon Microsoft yet relating to use of VR / TTS / SAPI.

Microsoft AI is coming; well I am guessing that you already know that.

Personally I would try both and utilize both and see how the shoe fits.

Here have turned down TTS over the years and never have been tethered to my cell phone.

[Timon]

I want to TTS however the TTS voice quality HS3 uses leaves much to be desire so so I haven't other than some tests with AirSpeak.

I'm taking a different track to get good TTS. Initially this only works if all the TTS you need is fixed speech not reading out information such as temperatures or other data.

Since I have an Amazon developer account I using Amazon Polly to create any speech files I need. You can create any speech phrase and then download the audio file. I'll then use Spuds AirSpeak to send the audio to any of the AirPlay speakers. That way you have really good speech quick and dirty.

However what I want to do is use Polly in its full glory . This way you send the text you want to Polly and have it send back the converted speech. The only downside to this is that you can get charged if you do a lot of speech conversion where downloading fixed speech is basically free.

What is the pricing? Basically the paid tier is $4 per one million characters of speech per month. That's close to 23 hours of speech. There is also a free tier that gives you five million characters per month for a year. Don't know what happens after the year is up.

Another way is to use both and cache the returned speech. You could download all of you're fixed TTS then only call Polly when you need to speak variable text.

The way I'm going to do it is to use cacheing. Caching will first check to see if you've spoken the phrase before and if so use the saved speech. If not, a call would be made to Polly and the returned file would be cached locally.

I think I've figured out an easy way to do this in a shell script and in a way that will not only work with AirSpeak but also with HS speaker client. If I get it to work the way I want I'll publish it under an open source license. However you'll need your own Amazon developer account to use it as I'm not setting up one for all HS3 users to use.

Since I've got other things to get worked out before I start this project I don't expect to get to this till next month.

BTW, is HS wanted to do this they could setup an account and add this to the myhs account. If using cacheing on both the local HS3 system and the HS interface to Amazon they likely wouldn't end up with a very large bill even with many people using the service.

[jjohnston]

Thanks Everyone! I think I'll build one mic with a respeaker and test. I was looking at the respeaker mic array v2 and also the Andrea Communications C1-1021450-100. Leaning more towards the ReSpeaker... ReSpeaker is $90 when you factor in the enclosure, the Andrea is $35...

[Rupp]

Thanks Everyone! I think I'll build one mic with a respeaker and test. I was looking at the respeaker mic array v2 and also the Andrea Communications C1-1021450-100. Leaning more towards the ReSpeaker... ReSpeaker is $90 when you factor in the enclosure, the Andrea is $35...

I would recommend spending some time in this forum dedicated to VR before proceeding as there have been some very elaborate systems built over the years and most of them were not as good as a $49 Amazon Dot.
https://forums.homeseer.com/forumdisplay.php?f=810

I would also gracefully disagree with Pete as my experience, over the years, is that Microsoft's Voice Recognition hasn't gotten any better and for some worse. In the XP days you could train the VR engine for a few hours and it seemed to work Ok with a very high quality array mic. Of course this was only for the user speaking. If your wife wanted to use it or the kids now they had to train the VR engine as well. Now the VR engine seems to require much more training. Either way good luck and be sure to let us know how this goes for you.

[Pete]

@Rupp,

I agree with you that the Amazon VR is excellent.

I have suggested to try both; cloud and cloud less speech and voice recognition.

[fresnoboy]

How big a deal is it that the VR stops working when the Internet is down? Are people really seeing that kind of unreliability? I hate Comcast, but's been very reliable at least in my area. And I can set up cellular backup if it was really that important.

For us, using the GH devices for VR isn't just about Homeseer, or asking random questions, it's about music and media in general. We have our google home devices set to play music through our amps using chromecast audio devices, and that works pretty well.

For HS to be the consolidated VR engine for the house, it has to do not just HS functions well, but music playback, and even video playback (using Flex TV plugin and Plex).

I don't know about Amazon, but Google is making the assistant and VR available for integration in lots of 3rd party devices. We have bestbuy versions of google home devices that have a clock display on them and make great alarm clocks too. HS could integrate the Google assistant in a more integrated way.

But for us, have multiple VR listeners in the home and teaching the family on what to say to each one is a nonstarter.

thx
mike

[Pete]

How big a deal is it that the VR stops working when the Internet is down?

The Amazon Alexa or GH will not work if there is no internet connection.

How big of a deal that is depends on you and your automation dependencies on the Amazon Alexa or GH.

Here primary ISP is Comcast with a failover to T-Mobile. Using PFSense failover works in a few seconds.

You can also integrate the Kinect and just gesture rather than speak to it.

It is cheap to test both these days.

Here also tinker with the cloud dependant Samsung Smartthings, Alexa integration and Roku integration.

But for us, have multiple VR listeners in the home and teaching the family on what to say to each one is a nonstarter.

There are a few Homeseer folks putting multiple Amazon Dot Listeners in the ceilings of their home these days.

Too bad they do not make a DOT with an ethernet connection for a combo POE in ceiling listener not connected wirelessly.

[fresnoboy]

Pete, I definitely prefer to keep the automation control local, and not in the cloud. That said, events, triggers and the rest can run on HS locally with no Internet access. It would be nice if HS clients on devices could also connect locally with no internet down, but the web frontend does work without internet support.

But I don't view VR as mission critical, so if it stops working but the automation system keeps running, that's an Ok tradeoff for me as long as it doesn't happen very often.

The quality and comprehensiveness of VR is pretty important though. My wife doesn't use Siri at all because of her initial experience with it. But she's fine with using the GH's.

thx
mike

[Pete]

It would be nice if HS clients on devices could also connect locally with no internet down, but the web frontend does work without internet support.

My 15 Wintel tabletop HSTouch tablets do not need the internet to work. The HSTouch that runs on the iPhone / Android can run locally via your WAP or via a VPN tunnel to your home if need be. In the Homeseer 1 and 2 days did open ports on my firewall for use external to my home network. Today do not do this and only utilize VPN to access my home network whatever device I want to get to.

I do a soup du jour cellular phone, not tethered and typically off when home.

[Blinder]

YEP!

To stay on topic, there is no comparison between the Microsoft voice recognition (which is what HS3 uses) vs Alexa/Google. The Alexa/Google VR is light years ahead of Microsofts VR to the point where the later will most like disappear in the next few years. Do yourself a favor and get a Dot and give it a go. Instant recognition with zero training.

I get why some might not want some cloud connectivity for mission critical stuff, but dam, I want all that horse power and knowledge that Google and Amazon provide. The only thing I have that's truly off-line for me is my DVR system that has 3 network cameras. That said, I also have a ring doorbell and flood light. Those video sources are cloud based.

I want $29 awesome devices that probably cost much more than that, but Google & Amazon need to make the devices invaluable so I suspect they are loss leaders or close to.

For those of us who are ok with the security implications and service issues, it's a no brainer what to use. I haven't had an IPS service outage in months (reset router) and haven't had a power outage in several years (no blinking microwave) My infrastructure is good to go.

I haven't used my Alexa much yet, but I have been using google home like crazy and I LOVE it. I don't want to talk to home seer, it's the back end. Alexa/Home are my verbal front end.

Just sayin

[gregb]

Cortana on Invoke speaker

Has anyone looked at using the Invoke speaker for voice recognition? A native implementation would be nice. I'd settle for IFTTT but I've been unable to figure out how the pass the text variable from the Cortana IFTTT service to the HomeSeer service.

With the Cortana IFTTT I can say Hey Cortana, turn on $ and like to pass the $ to HomeSeer so I only need on recipe for on and one for off without needing to create one for each device.