AI Phone (github.com)

A new AI phone agent, designed for macOS on Apple Silicon, allows users to make and receive calls through their computers, leveraging state-of-the-art local speech recognition and language models. The setup requires a minimum of 16 GB RAM for optimal performance with the 4B model and offers different configurations for various system capacities. Users can connect through SIP for direct calls, utilizing services like Telnyx or Twilio, while ensuring compatibility with available hardware. This announcement is significant for the AI/ML community as it deepens the integration of AI into real-world communication tools, enabling seamless voice interaction through locally running models that maintain user privacy. The architecture supports advanced voice synthesis via MLX, with capabilities for rudimentary call management and interaction control, making it an intriguing development for developers and researchers in voice technology. While currently limited to Apple Silicon and English, the project hints at broader applications and potential scalability across different platforms in the future.