TTS

[Abhrant]

This PR adds a TTS engine for e2e voice assistant on Pi.

TTS engine functions in the following way :

1. A separate thread having the tts process is called before the LLM :

  tts_processing_thread = threading.Thread(
                target=create_tts_wav,
                args=(stop_event, tts_processing_queue)
            )
  tts_processing_thread.start()

2. A queue of word outputs form the LLM is maintained and once a delimiter ['.', '!', '?' , ":" , ";"] is reached, queue is emptied and the sentence is sent to the STT engine.

3. STT used : https://github.com/rhasspy/piper

[Arnav0400]

Can we clean up the code (print statements, formatting, etc) and also have e2e benchmarks for this? Additionally can we also modify the config file to have audio out as a flag which triggers piper if set to true.

[Arnav0400]

can we have a relative path?

[Arnav0400]

lets have a relative path as default