Speech AI Assistant

An interactive AI assistant that can understand speech, generate responses, and speak back using text-to-speech.

Features

• Speech Recognition using Facebook's Wav2Vec2 model
• Text Generation using TinyLlama-1.1B-Chat
• Text-to-Speech using pyttsx3
• Real-time conversation capabilities

Requirements

• Python 3.12+
• Windows (for pyttsx3 text-to-speech)
• Internet connection (for downloading models)

Installation

1. Clone the repository:

git clone https://github.com/yourusername/speech_ai.git
cd speech_ai

2. Create a virtual environment:

python -m venv nemo_env

3. Activate the virtual environment:

# On Windows:
nemo_env\Scripts\activate

4. Install dependencies:

pip install numpy torch transformers pyaudio sounddevice pyttsx3

Usage

1. Make sure you're in the project directory with the virtual environment activated

2. Run the assistant:

python integrated_assistant.py

3. Once running:
   • Press Enter to start recording (speak for 5 seconds)
   • The system will:
     • Convert your speech to text
     • Generate an AI response
     • Speak the response back to you
   • Press 'q' and Enter to quit

Technical Details

• Speech Recognition: Uses Facebook's wav2vec2-base-960h model
• Text Generation: Uses TinyLlama-1.1B-Chat model
• Audio: Uses PyAudio for recording and pyttsx3 for text-to-speech
• Default recording duration: 5 seconds
• Sample rate: 16000 Hz

File Structure

• integrated_assistant.py: Main application file
• requirements.txt: Python package dependencies
• README.md: Documentation

Common Issues

1. PyAudio installation errors:

   • If you get an error installing PyAudio, download and install it from a precompiled wheel:
   • Visit PyAudio Wheels
   • Download the appropriate version for your Python
   • Install with: pip install PyAudio‑X.X.X‑cpXX‑cpXX‑win_amd64.whl

2. Text-to-Speech not working:

   • Ensure you have Windows SAPI5 voices installed
   • Check Windows Speech settings

License

MIT License