Voice is not quite understandable

[LtSnakePlissken]

Hey guys, I'm just wondering if anyone else is having issues understanding the voice? Is it possible to configure the voice somehow or use a different voice service? I'm not quite able to understand most of the time.

This thing is amazing by the way!

[judahpaul16]

Hey I'm glad you found the project useful!

In pyttsx3 on Linux, the default voices are provided by eSpeak, and they include a variety of languages and accents including female voices and even supports custom voice files.

Within common.py:

# ... engine = pyttsx3.init()

# Set the desired voice (example: setting American English voice)
engine.setProperty('voice', 'en-us')

# ... (rest of the file)

after modifying the file you may have to restart the app with gpt-restart or docker exec -it gpt-home supervisorctl restart app.

For more details, see the eSpeak documentation here and here.

If you want to train a totally custom voice like your own voice, you can check out TextyMcSpeechy by @domesticatedviking.

[judahpaul16]

Alternatively, you can opt to use a different TTS engine by replacing the pyttsx3 setup in common.py with your preferred TTS service.

# Initialize the text-to-speech engine
engine = pyttsx3.init()
# Set properties
engine.setProperty('rate', 145)
engine.setProperty('volume', 1.0)
# Direct audio to specific hardware
engine.setProperty('alsa_device', 'hw:Headphones,0')

You would invoke that service's version of engine.say() in the speak() function:

async def speak(text, stop_event=asyncio.Event()):
    async with speak_lock:
        loop = asyncio.get_running_loop()
        def _speak():
            engine.say(text)
            engine.runAndWait()
        await loop.run_in_executor(executor, _speak)
        stop_event.set()

Example using piper:

# Initialize Piper voice model and device
voice_model = 'your_voice_model'  # Replace with your specific voice model identifier
device = 'hw:Headphones,0'

async def speak(text, stop_event=asyncio.Event()):
    async with speak_lock:
        loop = asyncio.get_running_loop()
        def _speak():
            command = ['piper', '--voice', voice_model, '-o', device, '--text', text]
            subprocess.run(command)
        await loop.run_in_executor(executor, _speak)
        stop_event.set()