Improve voice control

[pnhofmann]

Since basic Voice control (#395) is merged (#408), I'm opening this as a discussion thread for any kind of improvements for voice control.

Any help or suggestions welcome!

[pnhofmann]

List of stuff, I would appreciate:

• Make speech recognition backend select-able (SpeechRecognition supports multiple backends
• Having a Speech recognition backend, which works offline
• Voice recorder without native requirements (e. g. using SoundCard, disadvantage: sometimes crashes)

[TobiOkanlawon]

I would like to work on this voice control feature. Please guide me as to where to begin, and also is it a good first issue. Kinda new to all this

[pnhofmann]

Hi!

and also is it a good first issue

Wouldn't know why not.

Please guide me as to where to begin

So obviously we don't write own Speech Recognition - but just use a library. In our case:

https://pypi.org/project/SpeechRecognition/

Voice control is implemented in jarvicli/plugins/voice_control.py. Guess you should just take a look and try to understand the code. Then you could just try bit experimenting, add some lines and check if everything behaves like you expected ;).

[TobiOkanlawon]

How about us implementing the offline voice control with pocketsphinx. If you are still particular about keeping the SpeechRecognition module then maybe you can implement something to check if there's internet connection and switch accordingly.

What do you think

[pnhofmann]

CMU Sphinx - which contains pocketsphinx - is in fact supported by SpeechRecognition. SpeechRecognition is a meta-module which unites multiple Speech Recognition services.

If you have a good reason to call pocketsphinx directly, that would perfectly ok. But yes: Having offline SpeechRecognition would be great and CMUSphinx was in fact what I had in mind.

[codecakes]

Was skimming through issues. Would like to know for which of the following backends, support is built?:

• CMU Sphinx (works offline)
• Google Speech Recognition
• Google Cloud Speech API
• Wit.ai
• Microsoft Bing Voice Recognition
• Houndify API
• IBM Speech to Text
• Snowboy Hotword Detection (works offline)

[pnhofmann]

Currently only Google Speech Recognition - Speech Recognition is in a really, really early stage...

[MikeyBeez]

Use vosk for STT. It's the best and all local.

[MikeyBeez]

https://github.com/MikeyBeez/Juliet

[pnhofmann]

Thanks for the tip! Will definitely take a close look at STT and Juliet

[gordodev]

Any update on this?

[pnhofmann]

Any update on this?

Let's hope soon ;).

[44-Unknown]

I working on jarvis project but I get error of Pyaudio how can I fix it

[pnhofmann]

@44-Unknown
What's does the error message say?

[MikeyBeez]

pip can't install pyaudio correctly. Use conda instead.

[Akul2010]

or install pyaudio wheel and pip install the wheel file name