Emotional AI Voice Chat

Fast conversation with emotional AI with even-realities G1 smart glass connection

This project implements a real-time ai conversation system with emotional text-to-speech (TTS) capabilities. It uses a large language model (LLM) for generating responses and a TTS engine with voice-cloning for voice output.

Based on:

• Voice TTS/STT implementation from LocalEmotionalAIVoiceChat
• LLM agent from memory-agent

Features

• Real-time speech-to-text input
• Cnversation generation powered by: Ollama, LMStudio, OpenAI, Anthropic or llama.cpp Webserver
• Emotion-aware realtime text-to-speech output
• Configurable system and user personas

Requirements

• Python <=3.10 (3.10.9 is recommended)
• Docker for redis and langraph

Installation

1. Clone the repository

git clone https://github.com/emingenc/G1_voice_ai_assistant.git
cd G1_voice_ai_assistant

2. Install the required packages

• venv (optional)

python -m venv venv
source venv/bin/activate

• Install the required packages

pip install -r requirements.txt

3. setup the environment variables in llm_agent folder example .env file

cp llm_agent/.env.example llm_agent/.env

Usage

Run the scripts in different terminals to start the system:

1. start the langraph

cd llm_agent
langgraph up
cd ..

2. start the redis

docker run --name my-redis -p 6379:6379 -d redis

3. start g1 smart glass connection

python g1_smart_glass.py

4. start the voice ai assistant

python voice_ai_assistant.py