Google stopped the TTS service, the PiAUISuite won't work

[r-hassan-zz]

Hello Steve,

It took 3 days to figure out tts script is not working.

Seems like google has blocked that service when its used by automated scripts. https://support.google.com/websearch/answer/86640

Can you please update like Jasper?

Thanks.

Raqueeb Hassan Bangladesh

[CoOral]

I've got exactly the same problem, and i'll be thankfull if you might update it.

Best regards

[r-hassan-zz]

Or you can use offline espeak tts script for rest of us.

I guess you had that previously.

[gymolnargabor]

yes, make it offline please... (and if it is not too difficult, can you remove the old videos from youtube, and make new ones for raspbian...it took me a lot of time to figure out, I'm a little bit outdated :) )

[shrutie]

@raqueebh How exactly can we use the offline espeak script? Please mention the changes you made, I am struggling since August.

[patrickreidglennon]

the tts script seems pretty basic, can't we just drop in another one and wrap it with a simplified /usr/bin/tts ? Going to give that a try

[patrickreidglennon]

Changes my tts script to this:

!/bin/bash string=$@

espeak -ven+f3 -k5 -s150 "$string"

I can now use tts "this is a test" but I am pretty sure voicecommand is returning null voice recognitions.. Having a look at /usr/bin/speech-recog.sh now, but I'm wondering if we need API keys now...?

[irsx02]

I have also noticed that google's TTS is no longer working. I downloaded pico2wave to replace tts.

I then wrote a wrapper to replace tts with this chunk of code

#!/bin/bash
#since google ended TTS, this script replaces tts with pico2wave.

if [ $# -lt 1 ]  
then               #an argument was not entered - pico2wave needs something to say
   /usr/bin/pico2wave -w /tmp/tempsound.wav "I have nothing to say." 
   /usr/bin/aplay -q  /tmp/tempsound.wav
   rm /tmp/tempsound.wav
   exit 0
fi
speech=$@  
/usr/bin/pico2wave -w /tmp/tempsound.wav "$speech"  
/usr/bin/aplay -q  /tmp/tempsound.wav
rm /tmp/tempsound.wav

What I found is that if I run "tts 'say something'" it will "say something." Now, when I'm running "voicecommand -c", I noticed that it will respond based on what's in the command.conf file. So, if i say "who made you" VoiceCommand will reply back with "I was created by Steven Hickson." However, the first acknowledge response, "Yes Sir" or the i-dont-understand-you prompt, "Received improper command" is not coming out at all.

I'm looking into /dev/shm/voice.log and such, and I think its something to do with VoiceCommand using "tts -l en" (where -l en for english ) vs pico2wave -l en-US, (where en-US represents US english.) Which means my script above is spitting out a -l error messsage.

So far, I've found speech-recog.sh that has a reference to "tts -l en" I fixedit to "tts -l en-US" but I'm still looking for other references.

[vaski666]

irsx02

where did you place the wrapper?

[Colin1964]

irsx02

So good to know someone is trying to keep voicecommand alive! Please could you provide a bit more detail on where this wrapper needs to go. Do you just replace the existing tts.sh file with this one?

[SweeneyToddTheDemonBarber]

Yes! Please keep trying. I've already installed Pico2Wave.

[irsx02]

Ok, I'm on Raspberry Pi 3, Raspbian (jessie), so your mileage may vary. I placed my TTS script at /usr/bin directory. There should already be a TTS file there. Back it up (you may need to use sudo), copy my tts script, and then make the script executable.
To test, type in tts hello world. my voice is back.

And you should hear its greeting.

[krist-jin]

well seems google tts change its api somehow.. But you can modify the part of calling tts api in /usr/bin/tts to fix: wget -q -U Mozilla -O output.mp3 "http://translate.google.com/translate_tts?ie=UTF-8&total=1&idx=0&textlen=32&client=tw-ob&q=Test&tl=En-gb" This works for me.

[irsx02]

Update; I made my tts replacement TTS script/wrapper a little more robust - but its now stuck at en-US (US English) It does solve the issue with the "Yes Sir" and "Receieved Improper Command." At the expense of behing locked into US English. Basically, I look for the -l option in tts, and make sure the language is set to "en-US" when it sees "en."

#!/bin/bash
#since google ended TTS, this wrapper-script replaces tts with pico2wave.
#version 0.2 -now rudimentarily handles language -l param.

if [ $# -lt 1 ]  
then                  #no argument entered - i need something to say
   /usr/bin/pico2wave -w /tmp/tempsound.wav "I have nothing to say." 
   /usr/bin/aplay -q /tmp/tempsound.wav
   rm /tmp/tempsound.wav
   exit 0
fi

if [ "$1" = "-l" ]    #-l in event where user explicitly defines language.  
then                   # Note: always assumes $2 is 'en' or a valid language option. 
   lang=$2
   if [ $lang = "en" ]   #TODO: cant find the real source of en, but if 
   then                      # i see 'en' I'm hard coding  en-US.
      lang="en-US"       #US English, mofo, do you speak it  
   fi           
   shift 2  
   speech=$@  
   /usr/bin/pico2wave -l $lang -w /tmp/tempsound.wav "$speech" 
   /usr/bin/aplay -q /tmp/tempsound.wav
   rm /tmp/tempsound.wav
   exit 0
else                  #else lets go straight to speech-output
  speech=$@  
  /usr/bin/pico2wave -w /tmp/tempsound.wav "$speech"  
  /usr/bin/aplay -q /tmp/tempsound.wav
  rm /tmp/tempsound.wav
fi 

I"m going to admit its not ideal, but it is working for me at the moment. It recognizes my keyword, respond appropriately (yes sir!) and awaits my command. When I tell it to "tell me the time" it replies correctly.

[Colin1964]

@irsx02

Firstly, much appreciate the extra detail and also the extra mile with the update.

I'm working with RpiB with Wheezy (I know I should be on Jessie but I'm being lazy and it has taken me days to get Voicecommand just to do STT). I've installed your replacement TTS script (all of 2 secs, nice'n'easy) but now I am struggling to find a working version of Pico2Wav - when I apt-get libttspico-utils I get package not available?? Searching and looking on other forums suggests I have to compile and build???? Or there is this pico2wave "like" option here - (https://github.com/ch3ll0v3k/picoPi2). If i use this and change refs in your script from "pico2wave" to "PicoPi2Wav" would that work?

Thanks again for responses so far - I'm determined to get this working!!!

[Colin1964]

So I followed the instructions for PicoPi2Wav and it failed to compile. Looks like I'm going to have to find the real pico2wav.

[Colin1964]

OK, making some progress - found this guide here http://rpihome.blogspot.co.uk/2015/02/installing-pico-tts.html. Just waiting for it to compile - fingers crossed!!

[Colin1964]

@irsx02

SUCCESS!!! Thank You! - She is talking to me again.

[saikrishna321]

@irsx02 i followed your steps, now i'm able hear back from pi when i type tts "hello i'm back" i can hear the same thing

when i run the command voicecommand -c i always hear No translation

can you help me on this ??

Opening config file...
You have a formatting error on line 31 of your config file. I'm ignoring that line
running in continuous mode
keyword duration is 2 and duration is 3
Found audio
Recording WAVE 'stdin' : Signed 16 bit Little Endian, Rate 16000 Hz, Stereo
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100 46331    0    14  100 46317      1   5818  0:00:07  0:00:07 --:--:-- 10476
No translation
Found audio
Recording WAVE 'stdin' : Signed 16 bit Little Endian, Rate 16000 Hz, Stereo
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100 45918    0    14  100 45904      5  19152  0:00:02  0:00:02 --:--:-- 19158
No translation

[saikrishna321]

@Colin1964 can you paste your tts file in gist that would help me

[saikrishna321]

@krist-jin i tried the way you have mentioned, still no luck its keeping on saying No translation.

When i say Pie, i get a reply as Yes,Sir but after that when i ask Who made you? or anything it says No translation

can someone pls help me on this ?

[irsx02]

@saikrishna321 - Can you go into the PiAUISuite/VoiceCommand directory and look for a shell script called speech-recog.sh ? Run the script ./speech-recog.sh and say a few words If everything goes right, it should return

Recording WAVE 'stdin' : Signed 16 bit Little Endian, Rate 16000 Hz, Stereo
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100 45918    0    14  100 45904      5  19152  0:00:02  0:00:02 --:--:-- 19158
<Whatever you said within 2 seconds>

If it still says "no translation", then type leafpad speech-recog.sh It'll bring up leafpad on the code in speech-recog.sh Look for near the bottom of the file for "arecord -D $hardware -f ...." That line is the command that records your voice (arecord), converts it to flac audio format (flac), send the audio to google (curl) and parses the results (sed/awk)

The no translation is happening somewhere within those lines of code. You can run each section individually to figure out where the "No Translation" is coming from.

[irsx02]

One more thing. You should see what is the formatting error on line 31 of your config file. Via a terminal window, from your home directory

leafpad .command.conf

You should see a list of all VoiceCommand acceptable commands. Jump to line 31 and look for the format error.

(All instructions done on Raspberry Pi 3, Raspbian (Jessie) . Your OS, file locations and text editor may vary. )

[saikrishna321]

@irsx02 thanks a lot for getting back on it :) i will give it a try and get back to you .

[irsx02]

@saikrishna321 Also, your issue is more about STT (Speech To Text) and not TTS (Text to Speech) Text to Speech is what pico2wave is trying to solve. Speech to Text is more about "No translation"." See if anything over in #55 can help you out.

And remember, if you find the solution, make sure you post it to #55 to help out anyone in the future.

[saikrishna321]

@irsx02 i did take look at #55 but nothing helped me from that thread..

[saikrishna321]

@irsx02

When i ran ./speech-recog.sh i cant hear back anything ..

./speech-recog.sh
Recording WAVE 'stdin' : Signed 16 bit Little Endian, Rate 16000 Hz, Mono
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100 66840    0    14  100 66826      1   8860  0:00:07  0:00:07 --:--:-- 16681

when i run voicecommand -c again, it replies Yes Sir. Fixed the error on the .command.conf

after that it says No translation

voicecommand -c
Opening config file...
running in continuous mode
keyword duration is 2 and duration is 3
Found audio
Recording WAVE 'stdin' : Signed 16 bit Little Endian, Rate 16000 Hz, Mono
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100 66388    0    14  100 66374      7  35797  0:00:01  0:00:01 --:--:-- 35800
No translation
Found audio
Recording WAVE 'stdin' : Signed 16 bit Little Endian, Rate 16000 Hz, Mono
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100 65817    0    14  100 65803      9  46479  0:00:01  0:00:01 --:--:-- 46471
No translation

speech-recog.sh

https://gist.github.com/saikrishna321/833e0206741dbe5f8caff7036d58b090

can you please help me fix this ..

[irsx02]

The "No translation" error looks exactly like the STT(speech to text) issue mentioned in #55. Follow @krist-jin instructions from his post.

From the gist link you provided, I can see the line near the end with ... -f cd -t wav -d ... This has been changed and @krist-jin / @ripleyXLR8 has a a post that describes how to fix it. Google Speech API is not recognizing multi-channel audio anymore. His instructions will describe how to fix it.

[saikrishna321]

@irsx02 @krist-jin @ripleyXLR8

i tried exactly as mentioned by you guys and still no luck ..

#!/bin/bash

hardware="plughw:1,0"
duration="3"
lang="en"
hw_bool=0
dur_bool=0
lang_bool=0
for var in "$@"
do
    if [ "$var" == "-D" ] ; then
        hw_bool=1
    elif [ "$var" == "-d" ] ; then
        dur_bool=1
    elif [ "$var" == "-l" ] ; then
        lang_bool=1
    elif [ $hw_bool == 1 ] ; then
        hw_bool=0
        hardware="$var"
    elif [ $dur_bool == 1 ] ; then
        dur_bool=0
        duration="$var"
    elif [ $lang_bool == 1 ] ; then
        lang_bool=0
        lang="$var"
    else
        echo "Invalid option, valid options are -D for hardware and -d for duration"
    fi
done

#this works really inconsistently and I don't know why. I would love to implement it
#sox -r 16000 -t alsa $hardware /dev/shm/out.flac silence 1 0.3 1% 1 0.5 1%
#wget -q -U "rate=16000" -O - --post-file /dev/shm/out.flac --header="Content-Type: audio/x-flac; rate=16000" "http://www.google.com/speech-api/v1/recognize?lang=en&client=Mozil$
#arecord -D $hardware -f cd -t wav -d $duration -r 16000 | flac - -f --best --sample-rate 16000 -o /dev/shm/out.flac 1>/dev/shm/voice.log 2>/dev/shm/voice.log; wget -O - -o /dev$
#arecord -D $hardware -f cd -t wav -d $duration -r 16000 | flac - -f --best --sample-rate 16000 -o /dev/shm/out.flac 1>/dev/shm/voice.log 2>/dev/shm/voice.log; curl -X POST --da$
arecord -D $hardware -f S16_LE -d $duration -r 16000 | flac - -f --best --sample-rate 16000 -o /dev/shm/out.flac 1>/dev/shm/voice.log 2>/dev/shm/voice.log; curl -X POST --data-b$
rm /dev/shm/out.flac

inline float GetVolume(string recordHW, string com_duration, bool nullout) {
FILE *cmd;
float vol = 0.0f;
string run = "arecord -D ";
run += recordHW;
run += "-f S16_LE -d";
run += com_duration;
run += " -r 16000 /dev/shm/noise.wav";
if(nullout) run += " 1>>/dev/shm/voice.log 2>>/dev/shm/voice.log";
system(run.c_str());
cmd = popen("sox /dev/shm/noise.wav -n stats -s 16 2>&1 | awk '/^Max\\ level/ {print $3}'","r");
fscanf(cmd,"%f",&vol); fclose(cmd);
return vol;
}

sudo apt-get install g++-4.8 make go to install folder and sudo ./InstallAUISuite.sh

now when i run voicecommand -c cant hear back anything, i had to change the file /usr/bin/tts as below to head back the voice, i cant hear anything other than Yes Sir?

wget -q -U Mozilla -O output.mp3 "http://translate.google.com/translate_tts?ie=UTF-8&total=1&idx=0&textlen=32&client=tw-ob&q=$tmp&tl=En-gb"

wget -q -U Mozilla -O output.mp3 "http://translate.google.com/translate_tts?ie=UTF-8&total=1&idx=0&textlen=32&client=tw-ob&q=$string&tl=En-gb"

[irsx02]

@saikrishna321 first thing, make sure you can record any sound on your mic. You may search google on how to use arecord / aplay to record and play saved audio.

After you confirmed your sound is working. Try this

(All lines of code here come from the /PiAUISuite/VoiceCommand/speech-recog.sh file, but I filled in any variables.)

First, record your voice and convert to flac. The command to do it is ;

arecord -D plughw:1,0 -f S16_LE -d 3 -r 16000 | flac - -f --best --sample-rate 16000 -o /dev/shm/out.flac 1>/dev/shm/voice.log 2>/dev/shm/voice.log

Now, go to /dev/shm directory and make sure there's a out.flac file present. If its there, run this command (this will send your speech up to google for translation, and get google's results)

curl -X POST --data-binary @/dev/shm/out.flac --user-agent 'Mozilla/5.0' --header 'Content-Type: audio/x-flac; rate=16000;' "https://www.google.com/speech-api/v2/recognize?output=json&lang=en-US&key=AIzaSyBOti4mM-6x9WDnZIjIeyEU21OpBXqWBgw&client=Mozilla/5.0" Look at the result. Its XML (i think) There should be a parameter called "transcript" followed by text of what you said.

Now, I think this is the STT part that isn't working. Please show me the results when you run curl. I'll help you troubleshoot it from there

[saikrishna321]

@irsx02 will give it a try today, once i have the PI with me :) Thanks a lot for detailed info

[saikrishna321]

@irsx02

i can see what i recorded in the XML file

pi@raspberrypi:/dev/shm $ curl -X POST --data-binary @/dev/shm/out.flac --user-agent 'Mozilla/5.0' --header 'Content-Type: audio/x-flac; rate=16000;' "https://www.google.com/speech-api/v2/recognize?output=json&lang=en-US&key=AIzaSyBOti4mM-6x9WDnZIjIeyEU21OpBXqWBgw&client=Mozilla/5.0"
{"result":[]}
{"result":[{"alternative":[{"transcript":"I am from India","confidence":0.77188921},{"transcript":"I'm from India"}],"final":true}],"result_index":0}

[saikrishna321]

@irsx02 ping

[irsx02]

@saikrishna321, I'm in California, USA. We're going to play time zone tag. Plus, I don't have my pi right now, and can't get to testing it until the evening, my timezone.

In the mean time, I see from your curl output that its able to contact google correctly. Curl command posts your audio file ("I am from India") up to google speech and turn it into text, which is the transcript response you posted. I'm thinking that somewhere in your copy of VoiceCommand, its not recording the audio correctly. I know #55 took care of my issue with "No Translation" When I get back I can see if I can find more info.

For now, make sure the cpp file mentioned in #55 is using the new code, and same with speech-recog.sh. I don't recall what else I did off the top of my head.

[saikrishna321]

@irsx02 hope you a great time out there..

with further debugging

I record using the below command

arecord -D plughw:1,0 -f S16_LE -d 3 -r 16000 | flac - -f --best --sample-rate 16000 -o /dev/shm/out.flac 1>/dev/shm/voice.log 2>/dev/shm/voice.log

Now i curl and i see XML response

pi@raspberrypi:/dev/shm $ curl -X POST --data-binary @/dev/shm/out.flac --user-agent 'Mozilla/5.0' --header 'Content-Type: audio/x-flac; rate=16000;' "https://www.google.com/speech-api/v2/recognize?output=json&lang=en-US&key=AIzaSyBOti4mM-6x9WDnZIjIeyEU21OpBXqWBgw&client=Mozilla/5.0"
{"result":[]}
{"result":[{"alternative":[{"transcript":"I am from India","confidence":0.77188921},{"transcript":"I'm from India"}],"final":true}],"result_index":0}

as we recorded the above response i tried to hear it back with aplay /dev/shm/out.flac i hear a old tv sound nothing other than that ..

Playing raw data '/dev/shm/out.flac' : Unsigned 8 bit, Rate 8000 Hz, Mono

[saikrishna321]

@irsx02 hope you had a great time out there.

with further debugging

when i run speech-recg.sh

speech-recog.sh
Recording WAVE 'stdin' : Signed 16 bit Little Endian, Rate 48000 Hz, Mono
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100 86900    0   140  100 86760     68  42183  0:00:02  0:00:02 --:--:-- 42198
"hello how are you doing"

When i did a ./InstallAUISuite.sh, during the setup i was able to hear back on the setup check but

Now when i run the voicecommand -c

voicecommand -c
Opening config file...
running in continuous mode
keyword duration is 2 and duration is 3

just hangs at this place ..

[irsx02]

@saikrishna321 I noticed that on my Raspberry Pi 3, when I run aplay /dev/shm/out.flac, it only returns static - also like the sound of an old TV. I assume its because aplay can not handle flac file format.

(Can any experts verify this? Can aplay understand the flac audio format? )

The good news is that you got speech-recog.sh to work. Based on your output, it is correctly changing your recorded voice into text. Which is great!

I'm not sure what's up with the voicecommand -c right now - I'll keep researching on it. Do you see any error messages showing up in the /dev/shm/voice.log file? (leafpad /dev/shm/voice.log)

As a side note, each time your install piAUISuite/VoiceCommand, you will have to fix the tts script. One way around it is that I've found if you directly edit the tts in the /PiAUISuite/VoiceCommand directory, it will install with the fixed tts every time.

[saikrishna321]

@irsx02

I will check the voice.log today and will keep you posted. When i changed the bit Rate to 48000 from 16000 and .flac to .wav then aplay was able to play the audio i recorded. But why voicecommand -c does keeps on researching not sure...

[saikrishna321]

@irsx02 After running the Update.sh, running the speech-rech.sh i can still the text and tts Hello talks back...

the log looks like this ..

running voicecommand -c, i dont hear back Yes Sir!

pi@raspberrypi:~/Documents/PiAUISuite/VoiceCommand $ voicecommand
Opening config file...
running in continuous mode
verifying command as well
keyword duration is 2 and duration is 2
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100 99702    0    14  100 99688      5  40505  0:00:02  0:00:02 --:--:-- 40523
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100 88735    0    14  100 88721      7  49235  0:00:01  0:00:01 --:--:-- 49234

Recording WAVE '/dev/shm/noise.wav' : Signed 16 bit Little Endian, Rate 48000 Hz, Mono

flac 1.3.0, Copyright (C) 2000-2009, 2011-2013  Josh Coalson & Xiph.Org Foundation
flac comes with ABSOLUTELY NO WARRANTY.  This is free software, and you are
welcome to redistribute it under certain conditions.  Type `flac' for details.

noise.wav: 73% complete, ratio=0.578^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^Hwrote 102228 bytes, ratio=0.532
Recording WAVE '/dev/shm/noise.wav' : Signed 16 bit Little Endian, Rate 48000 Hz, Mono
Recording WAVE '/dev/shm/noise.wav' : Signed 16 bit Little Endian, Rate 48000 Hz, Mono
Recording WAVE '/dev/shm/noise.wav' : Signed 16 bit Little Endian, Rate 48000 Hz, Mono
Recording WAVE '/dev/shm/noise.wav' : Signed 16 bit Little Endian, Rate 48000 Hz, Mono
Recording WAVE '/dev/shm/noise.wav' : Signed 16 bit Little Endian, Rate 48000 Hz, Mono
Recording WAVE '/dev/shm/noise.wav' : Signed 16 bit Little Endian, Rate 48000 Hz, Mono

flac 1.3.0, Copyright (C) 2000-2009, 2011-2013  Josh Coalson & Xiph.Org Foundation
flac comes with ABSOLUTELY NO WARRANTY.  This is free software, and you are
welcome to redistribute it under certain conditions.  Type `flac' for details.

noise.wav: 73% complete, ratio=0.498^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^Hwrote 93399 bytes, ratio=0.486
Recording WAVE '/dev/shm/noise.wav' : Signed 16 bit Little Endian, Rate 48000 Hz, Mono

flac 1.3.0, Copyright (C) 2000-2009, 2011-2013  Josh Coalson & Xiph.Org Foundation
flac comes with ABSOLUTELY NO WARRANTY.  This is free software, and you are
welcome to redistribute it under certain conditions.  Type `flac' for details.

noise.wav: 73% complete, ratio=0.478^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^Hwrote 87720 bytes, ratio=0.457
Recording WAVE '/dev/shm/noise.wav' : Signed 16 bit Little Endian, Rate 48000 Hz, Mono

[irsx02]

@saikrishna321
I noticed that in your voice.log, you are seeing

ALSA lib pcm.c:2239:(snd_pcm_open_noupdate) Unknown PCM plughw:1,0-f
arecord: main:722: audio open error: Invalid argument  

This means somewhere out in the code, there is a missing blank space between the device "plughw:1,0" and the parameter " -f" You will have to look exactly where in the code, but its suppose to say "plughw:1,0 -f " but instead its saying "plughw:1,0-f" (that lack of space between the 1,0 and -f is messing with the audio.

[irsx02]

One more thing to try out. In the voicecommand.cpp file, under the GetVolume(string recordHW, string com_duration, bool nullout) function, look for the line

run += "-f S16_LE -d";

I think its suppose to have a space between the " and the - f So it should look like this

run += " -f S16_LE -d ";

You will then need to run make (to compile the C++ code), install (AUISuite) and fix tts.sh again. Test this out and see if its the case.

[saikrishna321]

@irsx02 i did that and when i run voicecommand -c i see this below but dont hear anything

voicecommand
Opening config file...
running in continuous mode
verifying command as well
keyword duration is 2 and duration is 2
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100 82884    0    14  100 82870      8  48294  0:00:01  0:00:01 --:--:-- 48292
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100 80873    0    14  100 80859      1  11059  0:00:07  0:00:07 --:--:-- 16930

voice.logs

Recording WAVE '/dev/shm/noise.wav' : Signed 16 bit Little Endian, Rate 48000 Hz, Mono

flac 1.3.0, Copyright (C) 2000-2009, 2011-2013  Josh Coalson & Xiph.Org Foundation
flac comes with ABSOLUTELY NO WARRANTY.  This is free software, and you are
welcome to redistribute it under certain conditions.  Type `flac' for details.

noise.wav: 73% complete, ratio=0.451^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^Hwrote 82870 bytes, ratio=0.432
Recording WAVE '/dev/shm/noise.wav' : Signed 16 bit Little Endian, Rate 48000 Hz, Mono
Recording WAVE '/dev/shm/noise.wav' : Signed 16 bit Little Endian, Rate 48000 Hz, Mono
Recording WAVE '/dev/shm/noise.wav' : Signed 16 bit Little Endian, Rate 48000 Hz, Mono
Recording WAVE '/dev/shm/noise.wav' : Signed 16 bit Little Endian, Rate 48000 Hz, Mono
Recording WAVE '/dev/shm/noise.wav' : Signed 16 bit Little Endian, Rate 48000 Hz, Mono
Recording WAVE '/dev/shm/noise.wav' : Signed 16 bit Little Endian, Rate 48000 Hz, Mono

flac 1.3.0, Copyright (C) 2000-2009, 2011-2013  Josh Coalson & Xiph.Org Foundation
flac comes with ABSOLUTELY NO WARRANTY.  This is free software, and you are
welcome to redistribute it under certain conditions.  Type `flac' for details.

noise.wav: 73% complete, ratio=0.423^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^Hwrote 80859 bytes, ratio=0.421
Recording WAVE '/dev/shm/noise.wav' : Signed 16 bit Little Endian, Rate 48000 Hz, Mono
Recording WAVE '/dev/shm/noise.wav' : Signed 16 bit Little Endian, Rate 48000 Hz, Mono
Recording WAVE '/dev/shm/noise.wav' : Signed 16 bit Little Endian, Rate 48000 Hz, Mono

[irsx02]

@saikrishna321 I'm lost as to how to figure out this puzzle. I'm going to recommend delete and reinstall. Re-download all the files from github. Before running install, make the following changes (which I am using memory, so please do double-check first)

1. update PiAUISuite/VoiceCommand/tts with the code mentioned earlier in this post. Make sure Pico2Wave is already installed (which I'm sure you have already done so)
2. update PiAUISuite/VoiceCommand/speech-recog.sh old line -f cd -t wav with -f S16_LE -d as described in ripleyXLR8 post in #55.
3. update PiAUISuite/VoiceCommand/voicecommand.cpp function GetVolume(string RecordHW,...) Again, look for -f cd -t wav with -f S16_LE -d This again is described in more detail on #55.
4. install G++4.8 (which I'm sure you have done so already) run make. This is described by krist-jin in #55.
5. Now, install VoiceCommand
6. Testing time. a) First, confirm your mic works (arecord/aplay) b) run tts "your message here" - Speaker should say "your message here" c) run ./speech-recog.sh - whatever you say it two seconds will be output in text on screen.. d) run voicecommand -c - speak your response word, then after VoiceCommand says "yes sir" say "who made you" VoiceCommand should reply with Steven Hickson.

[saikrishna321]

@irsx02 Finally 👍 got it worked :) I followed the steps as you suggested, i had to change the channel to 32000 rate and it did work :) i can hear back ...

An other question, sorry --- Is there a way we can check google and get back for responses like Ok Google ?

[irsx02]

That's awesome - looks like a reinstall of VoiceCommand was all it took to get it to work.

As for the google / Ok google feature - I've been searching for something like that idea, but I have not found anything. The closest I did get (and by close, like India to California close) was to use the google function in VoiceCommand. I had to alter the google.sh script to use iceweasel to do a search. (I don't have midori)

Basically, I went to /usr/bin/google.sh and changed the code in it to call iceweasal.

Now, when I say "Google What is Raspberry Pi " VoiceCommand will open up a new tab in Ice Weasal witha search of "what is raspberry pi" No, its not exactly like "OK Google." But it is something.

[saikrishna321]

@irsx02 Will give it a try ... Take a look at this

https://oscarliang.com/raspberry-pi-voice-recognition-works-like-siri/

[irsx02]

@saikrishna321 Based on what I've read on this site ([ http://diyhacking.com/best-voice-recognition-software-for-raspberry-pi/]) and on Oscar Liang's site, both Oscar's and Voice Command 3.0 are very similar - uses google for stt, and google API for tts (although I had to change it over to Pico2Wave, which is owned by Google)

If you're familar with python, you may perfer Oscar's Voice Recoginition.

But so far I like how I am able to use Voice Command's command.conf file to write my own commands and what to execute. And I'm a little bit familar with C++

EIther way, neither of these two options have an "OK google"-like feature. That's what I am looking for.

[saikrishna321]

@irsx02 when you say neither of them have a feature like OK google-- i thought Oscar's Voice Recoginition has it right... let say when you ask OK Google-- What is 3 times 3 it return you 9, its the same with Oscar's Voice Recoginition. But with DIYHacking it just pulls from the command.conf and in return would need as execution statement ..

Just correct me if i'm wrong out there ?

[saikrishna321]

@irsx02 i see something weird, i notice once i run the voicecommand -c and ask for the first time who made you ;; it replies the right one.. Now it keeps on saying yes sir and waits for command.. i was expecting Yes Sir only when i call the keyword .. Can you please help me on this?

#This is the default config file
#These are the special options you can set (remove the #)
#!verify==1
!keyword==Sam
#!thresh==0.7
!continuous==1
!response==Yes Sir?
#!quiet==0
#!ignore==0
!filler==0
!duration==2
#!com_dur==3
#!hardware==plughw:1,0
#Here are the commands
show me==/home/pi/AUI/Imaging/test 2
track me==/home/pi/AUI/Imaging/test 1
download==download ...
play $1 season $2 episode $3==playvideo -s $2 -e $3 $1
download $1 season $2 episode $3==download $1 s$2e$3
play==playvideo -r -f ...
multiple==playvideo -r -m -c 5 ...
download==download ...
YouTube==youtube-search ...
Google==google ...
~music==xterm -e pianobar
~weather==python /home/pi/Documents/pisuite/weather.py
~made you==tts "I was created by Steven Hickson" 2>/dev/null
~music==xterm -e control-pianobar.sh play
~terminal==xterm &

[SweeneyToddTheDemonBarber]

Yes my software is having the same problem. Everything works perfectly except the keyword has 0 confidence. Any loud noise activates my keyword (which is Perkins), even clapping. I've already done what's written here and in #55 What can I do?

To add to this I figured out that the voicecommand is only recording the first letter of what I say. I was trying to change the commands in voicecommand -e and saw the $1 and $2. After I tell Perkins to Repeat $1==tts "$1" he only says one letter. The same happens with other commands.

I believe it's only recording the first sound and that is why the keyword always goes off.

[saikrishna321]

@SweeneyToddTheDemonBarber when you try ./speech-recong.sh say something now.. and see if it prints completely what u say