HomeSeer Message Board

HomeSeer Message Board (http://board.homeseer.com/index.php)
-   HomeSeer Voices (http://board.homeseer.com/forumdisplay.php?f=653)
-   -   Voices on Linux ? (http://board.homeseer.com/showthread.php?t=178868)

petez69 January 6th, 2016 06:58 PM

Voices on Linux ?
 
Ahh the voices and they arent in my head :-)

Do voices work with the Linux HS3 package ? I'm missing Allison and I dont really want to go back to Windows :-)

Thanks..pete

Pete January 6th, 2016 10:56 PM

1 Attachment(s)
Do voices work with the Linux HS3 package ?

yes. You can purchase voice fonts today for Linux today.

I have been able to get MS SAPI to work with Linux Wine. I did have to throw the kitchen sink to wine though.

I'm missing Allison and I dont really want to go back to Windows :-)

I am using windows for my Homeseer SAPI speech. Base tabletops run XPe with MS SAPI.

New test touchscreen is running Windows 10 (PipoX7), MS SAPI, HS Touch and Kinect & Amazon Alexa. Works fine.

I utilize MS Sapi Loquendo Australian Grace speech font running on a VB on the Ubuntu 14.04 machine.

Loquendo Grace

Pete January 7th, 2016 02:06 AM

BTW here is:

SAPI on Linux called Milena which looks easier than using Wine.

Here is a sample of SAPI running on Linux. (in polish)

SAPI Linux Sample

mikaluch March 31st, 2016 03:40 AM

Voice RSS
 
There was a post in the Howto's about using 'Voice RSS' - a web TTS service. But I can't post a question there. Registering to get the API Key has "Code protection" as a mandatory field in the form. I guess that must be a developer's secret term. Does someone know an appropriate entry?

jhearty May 22nd, 2016 04:53 PM

Ever figure this one out? I tried that too. I think I saw that field but left it blank. I got registered and got a key, but using it in the script from that other post all I get is silence.

mikaluch May 22nd, 2016 05:59 PM

Rmasonjr responded with a PM. Turns out code protection is the captcha. I haven't got to working on it again. The only speech interaction I'm doing right now is through Alexa and Echo. We can use the Echo and Dots, our phones with the FireTV remote app, the FireTV remote controls, and now Roger to communicate with Alexa. Adding more voice capabilities is low on the priority list.

I do have my old Kate and Paul NeoSpeech disks but so far have been unsuccessful at loading them on Windows 7, although honestly I haven't tried very hard. In fact the only thing that has worked well for me first try on that Windows 7 machine is iSpy. Unfortunately for iSpy, Zoneminder has also worked very well for me so I doubt I will be using it. Although iSpy appears to be pushing forward with facial recognition while Zoneminder is kind of poo pooing the concept even though they have a (broken) plugin.

My HomeSeer software integration has slowed recently because I'm spending a lot of time installing devices which requires fixing the jacked up wiring in my home. Whoever did the wiring here must have been color blind.

BlairG May 30th, 2016 04:47 PM

Better sound quality through Echo
 
Here I've got better sound quality by using a small Bluetooth Transmitter plugged into the headphone jack on the living room Android touch tablet.

The living room Echo isn't too far away and stays paired with the transmitter all the time. When Homeseer speaks or plays a sound file it comes out of the Echo speaker. Much better quality that the little speaker embedded in the tablet.

BigBadBuzz August 24th, 2016 06:26 PM

Cepstral
 
I've been able to get the swift speech engine provided with the Linux version of Cepstral voices to speak via Lubuntu command line and via a bash script; but when I put the same commands into the speak.sh script that I believe the Linux version of HS uses to "speak", I get nothing. The voices are not licensed but according to Cepstral, unlicensed voices work but have the "this is unlicensed" message prefixed to the beginning of the speech. Has anyone been able to get swift/Cepstral to work with HomeSeer? :confused:

BBB

BigBadBuzz August 31st, 2016 08:28 PM

Hopefully getting closer...
 
Ok -

I have a bit more information. By default, my HS3 is set to run as a service (via systemd) on my Linux server; it is this configuration that I have not been able to get swift/Cepstral working with HS3. The service script uses the Linux 'screen' command to leave the "window" that HomeSeer is running in alive (but detached). Tonight, I killed the HomeSeer service and started HS3 directly from the ./go script in /usr/local/HomeSeer. With HS3 running this way, the Cepstral TTS (I'm trying the Callie voice) works! SO there is something about HS3 running in a Linux 'screen' session preventing Cepstral's swift speech engine (command line) from working with HS3 - does anyone have any words of wisdom?

Regards,
BBB

Pete August 31st, 2016 11:04 PM

Good news there BBB!

Thinking you are the first person to purchase the switch/Cepstral TTS fonts for Linux that I have read about on the forum.

Disable the service and start HS3 as a cron job and see if that works for you.

Here still using my MS SAPI fonts running in Mini Windows server VM on the HS3 Ubuntu server. Using the same VM now for second speaker dot exe connected to the Zee-2 (HS3 Lite). Using a different voice font such that I can tell which one is talking.

Initially I was able to get MS SAPI in Wine but couldn't get speaker dot exe to run in Wine.

BigBadBuzz September 1st, 2016 06:02 PM

Pete:

Thanks for the suggestion - but it was no go. First I disabled the homeseer service I had set up via systemd. Then I edited root's crontab by typing

Code:

sudo crontab -e
and added a single line at the bottom of the file:

Code:

cd /usr/local/HomeSeer;./go
HomeSeer started when I rebooted the machine, but no Cepstral speech. I'll put in a ticket over at Cepstral to see if they can explain this, but I'm not overly confident I'll get a usable answer. I'd really like to not have to manually start HomeSeer every time I reboot the little server, but I do like having the nicer, less-robotic voices.

Regards,
BBB

BBB

BigBadBuzz September 1st, 2016 07:16 PM

Well, doesn't looklike Cepstral is the one...
 
All:

After the wrangling with Cepstral's TTS engine 'swift', I discovered today that even though it has the command options to send the output to a sound file (WAV file, by default), the personal voices (at least for Linux, but I think for Windows and MacOS too) are NOT licensed to do that - you have to buy the "server" versions. So, without shelling out an unknown amount of more dollars, I wouldn't be able to use Cepstral voices in the speak_to_file.sh script; I use it via the speaker client in the Sonos plugin. I really don't want to use one kind of TTS voice in some instances and something totally different in others. Oh well, I'll keep exploring...

Regards,
BBB

Pete September 2nd, 2016 04:48 AM

Thanks BBB. Well now we know about the Linux Cepstral voices.

A while back someone did write a Linux MS SAPI engine. I am not sure if the application is still around these days. By default here I use Wine on my laptops. There are some older MS programs that work fine in Wine. IE: here utilize PCA (Leviton HAI PC Access) and it works good enough for me.

One of the original things that attracted me to HS early on was the use of MS SAPI. Such that I purchased Homeseer's offerings back then of AT&T / Neospeech voice fonts. Neospeech seemed to be a bit lighter and nicer sounding that AT&T.

The mini Windows server VB on the Ubuntu 16.04 box works great for MS SAPI. Tested also to work fine was the Kinect / Alexa plugin talking via Wintel to the HS3 on the same box. The Alexa voice does work fine and is distributed via the Russound audio.

I did purchase a bunch of Kinects to test with. Too bad that MS doesn't make a miniature wireless Kinect.

Today too all of my 15+ HSTouch embedded touch clients do MS SAPI and speaker dot exe such that it the client can just be a Homeseer SAPI TTS zoned device. One client I use in the master bedroom streams audio via push button on the touch client while running speaker dot exe such that I hear HS announcements concurrent with my streaming audio.

Personally here never have HS plugin proxied the HS TTS speech. I have mixed the audio stuff with the old AB8SS / Russound system but never have put dependencies on HS for my multimedia. IE: Omnitouch screens / HSTouch do control the Russound system but fastest for me is the Russound keypads. Over the years have toned down the audio a bit. After sleep time there is no more TTS audio. Sensors only trigger chimes these days after hours. I can though just press a button on the tabletop and have it tell me what exactly was the cause of the trigger chimes.

The zoned russound audio here is sub zoned in a few areas (well all of the bedrooms for one). Master bedroom has 5.1 sound but 2.0 zoned audio if that makes sense? I have a HSTouch client tabletop connected via a digital output to the receiver and a simple AB switch to change the MB / master bath audio from zoned audio to local audio.

BigBadBuzz September 2nd, 2016 05:04 PM

Pete:

Thanks for the information. Myself and two buddies all started using HomeSeer at the same time, and I'm the "odd man out" as the one using Linux - so I get some good-natured "grief" when Linux doesn't support some of the things that Windows does. I would have preferred to have stayed with just Linux and not run a VM, but the SAPI capabilities / voices and the ability to run the program that interfaces to Alexa voice synthesis (I think it is only available for Windows) are convincing me to give VirtualBox a try. What version of Windows do you run in VirtualBox? Is there a way to trim down the Windows installation to minimize the resources it uses?

I'm trying the speaker client in the Sonos plugin because the Sonos boxes are always on, and so far as a speaker client it seems to work pretty well for announcements - it automatically ramps down the volume of what was playing to 0, speaks the announcement, then ramps the volume back up to what it was and starts playing the same source (if a source was playing). Otherwise I use the Sonos plugin just for on/off and volume control via Alexa. I like how you use chimes at night - I'll give that a try.

Regards,
Paul (BBB)

RandyInLA September 5th, 2016 06:23 PM

Questions about MS SAPI & VR for Rpi3
 
So on a Windows HS setup, MS SAPI handles both voice recognition AND text to speech? And the way to access MS SAPI's VR from within HS is through the speaker.exe program...which won't run on my Rpi3?

I'm running HS .297 Linux version on Rpi3 via mono v4.4.2. For TTS, I was able to get Ivona working via Python scripting outside of HS. The English/British voice, Brian, sounds very much like Jarvis from the Ironman movies. I then tweaked the HS speak.sh script to call a python script to send text to Ivona. The returned .ogg file from Ivona is then played out the headphone jack on the Rpi3. I've never used or seen a speaker.exe program.
  • Even though I have TTS working, do I still need MS SAPI for VR?
  • Is there another VR engine I could use on the Pi, or is the HS voice scripting made specifically for MS SAPI?
  • If I setup MS SAPI on another machine (windows or windows in a VB) will it show up under "speaker clients" in the HS Tools/Setup/Network tab?
  • If it DOES show up, how do I get speaker.exe to run on the Pi? (or do I even need that app?)
  • I had a look at Milena SAPI and it seems to only cover TTS, not VR?
  • Pete, do you have your VB running on the same machine as HS? I assume you are running a full sized computer with Linux, and not an RPi3?
  • If I do figure out a way to get MS SAPI running for VR, will it then take over my TTS as well?!

How I got Ivona to do TTS on the Rpi3:
  • Create an Ivona.com account
  • Install Pyvona for Python
  • Create the following python script in the root dir of HS, named speak.py, to send text to Ivona and receive an .ogg audio file back:
    Code:

    import pyvona, sys

    v = pyvona.create_voice('your_access_key', 'your_secret_key')
    v.voice_name='Brian'
    v.speech_rate='medium'
    v.fetch_voice(sys.argv[1], 'temp')

  • Change the HS speak.sh script so it uses the speak.py script you just created, then plays the resulting .ogg audio file:
    Code:

    #!/bin/sh
    python speak.py "$1"
    omxplayer temp.ogg

  • Install the omxplayer program. The default audio playback app HS used wouldn't play .ogg files.
    Code:

    $ sudo apt-get omxplayer
Even though I just need the one text argument to create the audio file, I am in the process of trying to pass in a 2nd or even 3rd agrument to select which voice to render and/or which speakers to play out of. Will eventually use the Gentner XAP 800 audio matrix.

Pete September 6th, 2016 10:27 AM

Alexa voice synthesis and Kinect run fine on my PipoX7 configuration with a 15" multitouch screen talking to the Homeseer 3 mothership running in Ubuntu 16.04 64 bit.

The above will also run on a VM on the Ubuntu mothership. Here you would extend the Kinect over to the mothership.

What version of Windows do you run in VirtualBox? Is there a way to trim down the Windows installation to minimize the resources it uses?

Skinny Windows 2003 server, XPe, W7 and W10. Really you would only run what you need and keep it simple.

Is there a way to trim down the Windows installation to minimize the resources it uses?

W7 embedded and XP embedded can run fine with about 4Gb of OS space. I have Jogglers today running XPe on 8 GB or 16Gb SSDs just fine.

IE: skinny server is running SAPI and two speaker dot exe's from two HS servers.

Here have never utilized any proxy for TTS. That said have tested and used the Squeezebox plugin (not sure if it works in Linux today). My main sound if via a Russound zoned audio system (12 zones) plus still using the older AB8SS (8 zones) here.

I like how you use chimes at night - I'll give that a try.

I have two bedside touchscreens in which the screens are off at night. I can also play back sound; like a chime and just enable an alert screen such that if you touch it you will see it or if it wakes up the screen while chiming you can see it.

Usually one or two chimes after hours are attributed to animals or teens (no adult supervision - bad thing) wandering around at 3AM outside. Typically will check in the morning to see.

Pete September 6th, 2016 10:40 AM

on a Windows HS setup, MS SAPI handles both voice recognition AND text to speech?

yes

And the way to access MS SAPI's VR from within HS is through the speaker.exe program...which won't run on my Rpi3?

Yes MS SAPI doesn't run in Linux or on an RPi2. Only Windows and doesn't need the cloud to run.

Even though I have TTS working, do I still need MS SAPI for VR?

You can try VR in Linux. I haven't played with it since the 1990's and never tried it with Homeseer in Linux. I like my collection of MS SAPI fonts and continue to use them today.

Pete, do you have your VB running on the same machine as HS? I assume you are running a full sized computer with Linux, and not an RPi3?

Yes using an iSeries / 16Gb RAM PC mITX board for HS3 Pro. It is running Ubuntu 16.04 / 64 bit and a couple of VB's.

I did turn my old Zee-2 in to a HS3 slave device and moved my HS3 lite over to an Xi3/Xi5A dual core AMD micro PC with 2Gb of RAM. It will probably not be able to run VM's as 2Gb of RAM is a bit light.

If I do figure out a way to get MS SAPI running for VR, will it then take over my TTS as well?

Yup. MS SAPI running in a VB is using speaker dot exe. I am also running speaker dot exe on my touchscreens such that each instance can be a different voice and each can do VR locally.

I have an Amazon Echo today but not using it with Homeseer at this time. Rather testing Kinect and the Alexa plugin.

RandyInLA September 6th, 2016 12:46 PM

Quote:

Originally Posted by Pete (Post 1263309)
Even though I have TTS working, do I still need MS SAPI for VR?

You can try VR in Linux. I haven't played with it since the 1990's and never tried it with Homeseer in Linux. I like my collection of MS SAPI fonts and continue to use them today.

...

If I do figure out a way to get MS SAPI running for VR, will it then take over my TTS as well?

Yup. MS SAPI running in a VB is using speaker dot exe. I am also running speaker dot exe on my touchscreens such that each instance can be a different voice and each can do VR locally.

So if I spend the time/energy to configure a new windows-based machine with HS just so I can have access to voice recognition via MS SAPI, I then lose my ability to use the remote Ivona text to speech I am using now? I wonder why VR & TTS are locked together like that?! Ugh! I love the TTS I have working now and only want to add the ability to give voice commands.

If one were to try to use another Voice Recognition framework, how, exactly, would it interface with HS? Does HS simply take text output from the VR as text input? I'll have to have a look at the Alexa plugin you mentioned. I have the homebridge-homeseer config.json file working great and can control my devices with Siri, but I don't get the sort of interaction that is possible with HS VR.

Re:HS Touch and TTS. I noticed last night in the log that the HS Touch app on my iPhone was listed as a remote speaker whenever I launched it. So I had HS speak a phrase from an event... turned the speaker switch on on the iPhone, and it worked perfectly out of the iPhone speaker. Was some robotic female voice. At the same time, I also heard my Ivona voice play out the Rpi3 headphone jack, even though I specifically set HOMESEER:RANDY as the speaker client for the event. This makes me think I might still have access to my Ivona TTS even if I get MS SAPI working for Voice Recognition on another machine... hmmmm.... But do I want to spend the time/energy to get MS SAPI working remotely? Sigh. I probably will.

Thanks, Pete! A fine Tuesday to ye.

Pete September 6th, 2016 01:37 PM

I do not know what to write here Randy.

Microsoft has been developing it's SAPI TTS / VR framework for a very long time.

Here did play a bit with TTS on early computers that I had in the 1980's. (well late 1970's). Computers did start out as a hobby for me.

Fast forward to the 1990's and Windows 95. Used it then. Also purchased Dragon Speaking around that time.

Have a read here about the history of Microsoft Speech ==> Microsoft Basic architecture

First Homeseer box running first/second version of HS had text to speech and I used it.

Did also start playing with it with telephone banking in the 1990's. It was a special project of sorts and it did work at the time.

Just a couple of days ago played with VR on Windows 10. It is levels above my VR on embedded XP which works fine for me. Wish that a micro Kinect device was made.

If one were to try to use another Voice Recognition framework, how, exactly, would it interface with HS?

I have no clue.

Homeseer 3 for Linux works fine using Mono. There have been efforts to port over MS SAPI to Linux. I did get it to work in Wine. I just couldn't get speaker dot exe working in wine.

Been playing with cellular phones since the 1990's. Always played / always purchased phones unlocked and have never tethered myself to any phone or number. That is me.

You can purchase a mini pc for less than $100 with Windows 10 and use it for your speech TTS / VR stuff if you want.
  • Even though I have TTS working, do I still need MS SAPI for VR?

I dunno relating to keeping TTS/VR at home with Homeseer. You can utilize the Amazon Echo if you want. VR works fine with it. The Alexa app does give me what I want from Amazon.

I am getting lost in your thread somewhat.

Here just writing that I am using MS SAPI for Homeseer as I have now since the first version. I made it work for me using the original speaker dot exe Homeseer application with uses MS SAPI. I can run it today via a VM on my Homeseer 3 mothership. I have over the years tested out HSTouch on iOS, Android and MS. I only do MS today for Homeseer touch. The HS clients today run embedded XP. They do fine with SAPI and HSTouch. (IE: all 15 clients online 24/7).

RandyInLA September 6th, 2016 04:16 PM

Thanks, Pete. I thought about installing Wine on the same Rpi3 I have Homeseer running on. But as you said, you could get SAPI working on Wine but not speaker dot exe (why do you write it out "speaker dot exe" instead of speaker.exe?).

I'll spend more time searching through the forum for Voice Recognition ideas.


All times are GMT -4. The time now is 04:07 AM.


Copyright HomeSeer Technologies, LLC