/chatgpt-voice-assistant

A chatbot that integrates OpenAI Whisper, Chat Completions and Voice Generation. Also provides the option to use free transcription / TTS options.

Primary LanguagePython

ChatGPT Voice Assistant

GitHub Actions Build Status

A simple interface to the OpenAI ChatGPT model with speech to text for input and text to speech for the output. chatgpt-voice-assistant uses OpenAI Whisper for speech transcription and OpenAI Speech Generation for TTS.

Setup

Mac Prerequisites

Install dependencies:

brew install portaudio
brew link portaudio

Update your pydistutils config file for portaudio usage by running the following:

echo "[build_ext]" >> $HOME/.pydistutils.cfg
echo "include_dirs="`brew --prefix portaudio`"/include/" >> $HOME/.pydistutils.cfg
echo "library_dirs="`brew --prefix portaudio`"/lib/" >> $HOME/.pydistutils.cfg

Install from PyPI

Run the following to install the chatgpt-assist CLI application:

pip install chatgpt-voice-assistant

Install from Source

  1. Install poetry (official docs or with pip install poetry)
  2. Install all dependencies with poetry install

Running the Script

Either set the OPENAI_API_KEY environment variable before running the script or pass in your secret key to the script like in the example below:

export OPENAI_API_KEY=<OPEN API SECRET KEY HERE>
gptassist

# OR

gptassist --open-ai-key=<OPEN API SECRET KEY HERE>

or if installed from source with poetry:

poetry run gptassist --open-ai-key=<OPEN API SECRET KEY HERE>

Start speaking and turn up your volume to hear the AI assistant respond.

Say the word "exit" or hit Ctrl+C in your terminal to stop the application.

Options

Below is the help menu from the chatgpt-assist CLI detailing all available options:

-h, --help
    show this help message and exit

--log-level LOG_LEVEL
    Whether to print at the debug level or not.

--input-device-name INPUT_DEVICE_NAME
    The input device name.

--lang LANG
    The language to listen for when running speech to text (ex. en or fr).

--max-tokens MAX_TOKENS
    Max OpenAI completion tokens to use for text generation.

--tld TLD
    Top level domain (ex. com or com.au).

--safe-word SAFE_WORD
    Word to speak to exit the application.

--wake-word WAKE_WORD
    (Optional) Word to trigger a response.

--open-ai-key OPEN_AI_KEY
    Required. Open AI Secret Key (or set OPENAI_API_KEY environment variable)

--tts {apple,google,openai}
    Choose a text-to-speech engine.

--speech-rate SPEECH_RATE
    The rate at which to play speech. 1.0=normal

Specifying an Output Language Accent

Specify both the LANGUAGE and TOP_LEVEL_DOMAIN vars to override the default English (United States)

gptassist --open-ai-key=<OPENAI_KEY> --lang=en --tld=com

Language Examples

  • English (United States) DEFAULT
    • LANGUAGE=en TOP_LEVEL_DOMAIN=com
  • English (Australia)
    • LANGUAGE=en TOP_LEVEL_DOMAIN=com.au
  • English (India)
    • LANGUAGE=en TOP_LEVEL_DOMAIN=co.in
  • French (France)
    • LANGUAGE=fr TOP_LEVEL_DOMAIN=fr

See Localized 'accents' section on gTTS docs for more information

References