RSHVR/Command_RTC

Voice Chat Assistant

A conversational voice assistant powered by AI that responds to your spoken queries with natural-sounding speech.

Features

• —Speech Recognition: Uses OpenAI's Whisper model to accurately transcribe your voice
• —Natural Language Understanding: Leverages Cohere's LLM API for intelligent responses
• —Text-to-Speech: Generates natural speech using Chatterbox-TTS
• —Reply on Pause: Automatically responds when you finish speaking
• —Conversation History: Maintains context throughout your dialogue

Demo

Speak into your microphone and the assistant will respond with voice!

How It Works

• —Your voice is transcribed to text using Whisper
• —The text is processed by Cohere's LLM to generate a response
• —The response is converted to speech using Chatterbox-TTS
• —The conversation continues with full context retention

Technical Details

This project utilizes:

• —Zero-GPU: Efficient GPU memory usage with Hugging Face's Zero-GPU technology
• —FastRTC: Real-time communication for seamless voice interaction
• —Gradio: Simple and intuitive user interface

Setup

To run this locally, you'll need a Cohere API key and Python 3.8+.

Acknowledgements

• —OpenAI for the Whisper speech recognition model
• —Cohere for the language model API
• —Tortoise-TTS for the text-to-speech capabilities
• —Hugging Face for the Spaces and Zero-GPU infrastructure