JoeJoe

Voice activated Assistant

Voice Assistant with Whisper and Gemma

This project implements a voice assistant using OpenAI's Whisper for speech recognition and Ollama's Gemma model for natural language processing.

Features

• Wake word detection ("Hey Assistant")
• Speech-to-text conversion using Whisper
• Natural language processing using Gemma:2b
• Text-to-speech output
• Non-blocking response output and simultaneous threaded speaking

Requirements

• Python 3.9+
• Ollama (with Gemma:2b model)
• OpenAI Whisper
• SpeechRecognition
• PyAudio
• pyttsx3 (for non-macOS systems)

Installation

1. Clone this repository:

   git clone https://github.com/MikeyBeez/JoeJoe.git
   cd JoeJoe
   

2. Create and activate a Conda environment:

   conda create -n voice_assistant python=3.9
   conda activate voice_assistant
   

3. Install required packages:

   conda install -c conda-forge speechrecognition
   conda install -c conda-forge pyttsx3
   pip install ollama openai-whisper pyaudio
   

4. Install Ollama (See Ollama.com for instructions) and pull the Gemma:2b model:

   ollama pull gemma:2b
   

Usage

1. Activate the Conda environment:

   conda activate voice_assistant
   

2. Run the script:

   python main.py
   

3. Say "jarvis" to wake up the assistant, then speak your query.

Note

• Ensure your microphone is properly set up and recognized by your system.
• The first run may take some time as it downloads the Whisper models.
• This project runs Whisper on CPU. For better performance, consider using a GPU if available.

License

MIT