search

shivasiddharth / GassistPi

Google Assistant for Single Board Computers
Other
1.02k stars 303 forks source link

Change Voice does not change the hard-coded comments #690

Closed V1JCM closed 5 years ago

V1JCM commented 5 years ago

IMPORTANT NOTICE
If you do not complete the template below it is likely that your issue will not be addressed. When providing information about your issue please be as extensive as possible so that it can be solved by as little as possible responses.

FAILURE TO COMPLETE THE REQUESTED INFORMATION WILL RESULT IN YOUR ISSUE BEING CLOSED

Board and OS details:

Open a terminal and run 

cat /proc/cpuinfo  
cat /etc/os-release  
Paste your Board and OS details here.

Describe the bug:

Chose not to fill in the form, I guess the question is more generic. Can't find the answer from git. Goes like this:
1) "Hey google, change voice"
- Google changes voice, say to a male voice
2) Use GassistPi command like: "Hey Google, play I got you from Youtube"
- Google says "Getting autoplay links" in a default/female voice

Is there something to be done?

Did you go through or search the issues section to check if your issue was already discussed (either currently open issues or closed issues).

Yes/No

2. Is the issue related to assistant installation ?

Yes/No

If answer to question 2 is Yes, then paste the contents of the terminal below.

If answer to question 2 is No, then proceed further.

3. Does the assistant service start normally?

Yes/No

4. Is the assistant service restarting automatically?

Yes/No

If answer to question 4 is Yes, then paste the contents of the terminal below.

If answer to question 4 is No, then proceed further.

Manually start the assistant. For guidelines to start the assistant manually refer this

Paste the command below, that crashed the assistant

Paste the contents of the terminal below

shivasiddharth commented 5 years ago

The voice that says "Playing... etc etc" for custom commands is the voice of text to speech converter. That voice is not changeable with the current API that i am using. But an option to change the voice for custom commands is coming in the next update (if time permits mostly this weekend). You will not get variations like the google's voice but you can choose between male and female.

V1JCM commented 5 years ago

OK, great! No rush for our part, just wondering if anything can be done.

shivasiddharth commented 5 years ago

Replace your existing files with the new files from master. Open a terminal and run:

sudo apt-get install espeak
source env/bin/activate
pip install pyttsx3

Now you should be able to change the voice to male or female in the config.yaml. No need to re-run the installer. Backup your config.yaml file and may be you can delete the GassistPi folder and re-run the gitclone command alone.

V1JCM commented 5 years ago

Testing it, the voice is somewhat mechanical. And the startup "Hi!" comes with a different voice. Do you have any plans to update the API?

shivasiddharth commented 5 years ago

'Hi' is google's voice. I generated an audio file by recording my Google Home say "Hi". This new voice is the updated API. Old API did not have voice modulation capabilities. So you are saying that when the male voice is opted, even the startup tone should be male ? Sounds reasonable.

V1JCM commented 5 years ago

Oh, I see. It is no biggie, really. I could make my own mp3 as well, maybe something Jarvis would say.

I have done some more tests, there really is something wrong with my default voice. The "Google male" voice is ok, sounds very human. But the "No, I am forever alone" -voice is almost uncomprehensive. Sounds like something an old broken toy. Do you have any adice?

shivasiddharth commented 5 years ago

You should create a .wav file not .mp3 and it should be mono. I would suggest you to import the provided .wav files to some audio editor tool to check for the formats.

I am taking tiny steps to iron out the custom audio thingy. Previously gtts was used which had different tone, modulation and voice for each language (some languages will have a male voice and some a female one). The broad idea is to first create a voice with the same modulation for all the languages and then use a plugin to change the pitch to match a female or male voice.

By changing the API last night, i have managed to get a uniform voice for the different languages. Next step would be to alter its picth. Thats work in progress.

V1JCM commented 5 years ago

ok, cool. Happy coding!

shivasiddharth commented 5 years ago

I am receiving 👎s for the new voice from the users. On a scale of 1-10 how bad is it when compared to the old non-male voice? I really need inputs, if the new voice is that bad, then i will have to change direction for further work.

V1JCM commented 5 years ago

Don't want to be rude mate, please do not get offended. You are doing a great job for us all. But frankly, the voice is 2/10. Especially when you compare it to Google male (or female) voices, they sound quite human.

shivasiddharth commented 5 years ago

No issue mate. Feedback always helps. The entire project has been built and modified only based on users' feedbacks and perspectives.

shivasiddharth commented 5 years ago

Sorry to bother you again. Please find attached some male and female voices does this sound better ? The ones marked suffixed with "p" are the male versions of their counterparts. Audio Male and Female.zip

V1JCM commented 5 years ago

Happy to help! These are all better than the currently playing voice.

1 did not play, 5p.wav is my favourite of male voices and 6.wav of female. Both #2 are also good. #3 both sound like late professor Hawking. 4.wav female is ok, but #4 male sounds like he has too much vibration.

shivasiddharth commented 5 years ago

Robotic voice have been fixed. The voices now should be much better and if a male voice is chosen, now startup tone will also be male.

open a terminal and run:

sudo apt-get install sox
sudo apt-get install libsox-fmt-mp3

After that, rename your existing directory and re-clone the git.

V1JCM commented 5 years ago

OK, will do tonight after work. I'll report back to you then.

V1JCM commented 5 years ago

Thank You!

V1JCM commented 5 years ago

Did: sudo apt-get install sox sudo apt-get install libsox-fmt-mp3 Rename old GassistPI folder and: git clone https://github.com/shivasiddharth/GassistPi then copied API key from config.yaml to the new GassistPI Something is wrong as the service will not run. Manual start will cause all this:

/home/pi/env/lib/python3.5/site-packages/gmusicapi/init.py:4: GmusicapiWarning: gmusicapi.clients.OAUTH_FILEPATH is deprecated and will be removed; use Musicmanager.OAUTH_FILEPATH from gmusicapi.clients import Webclient, Musicmanager, Mobileclient INFO:gmusicapi.Mobileclient1:!-- begin debug log --! INFO:gmusicapi.Mobileclient1:version: 11.1.1 INFO:gmusicapi.Mobileclient1:logging to: /home/pi/.cache/gmusicapi/log/gmusicapi.log INFO:gmusicapi.Mobileclient1:initialized INFO:gmusicapi.Mobileclient1:logged out

:2: GmusicapiWarning: login is deprecated and may break unexpectedly; prefer Mobileclient.oauth_login INFO:gmusicapi.Mobileclient1:failed to authenticate Traceback (most recent call last): File "/home/pi/GassistPi/src/main.py", line 44, in from actions import say File "/home/pi/GassistPi/src/actions.py", line 19, in from langdetect import detect ImportError: No module named 'langdetect'
shivasiddharth commented 5 years ago

I just updated the project again with a new addon thingy. Run:

source env/bin/activate
pip install langdetect

Now things should start normally.

V1JCM commented 5 years ago

Seems to work now, thanks. Only it takes like 4-5 seconds to answer the "do you have a family" -question. And the speech volume needs to be set too.

Just a development idea: what about giving the user several voice options instad of just male and female?

V1JCM commented 5 years ago

This project has a different approach(?): https://dronebotworkshop.com/hacking-google-aiy-voice-kit-2/ I had that for a while, and I believe it used google speech as such to say custom frases. Like here: " def power_off_pi(): aiy.audio.say('Good bye!') subprocess.call('sudo shutdown now', shell=True) "

Would it be possible to use that approach in your project?

shivasiddharth commented 5 years ago

AIY uses pico2wave not cloud speech. Pico2wave has limited language support. It does not support Swedish, Dutch etc. With Google assistant available in a number of languages and more to come, I felt that gtts is more viable option for the long run of the project. Of course I am always on the lookout for alternatives, if some package better than gtts comes up, we will migrate to that.

V1JCM commented 5 years ago

OK, I see. Great thing about all this ICT we do is that it is never ready, never finished. Always developing, always going forward.

Thank you for your support, You are doing a great job!