Live commentary has been made possible by combining LiveKit for real-time video processing, Cerebrium for scalable AI infrastructure, and Cartesia for expressive voice synthesis. An AI commentator that can react and narrate in real time was built using the mentioned technologies. The system uses a queue to process frames from LiveKit, sending them to the model endpoint on Cerebrium for analysis and generating corresponding text and audio responses with Cartesia. The frontend handles video playback and publishing tracks to the room for streaming. For deployment, a .env file is created with necessary credentials and deployed to Cerebrium, while a separate LiveKit worker app is set up for scalable processing of video frames. The system aims to provide real-time commentary with minimal latency, rivaling human broadcasters, and opens up possibilities for further innovation in the future of live broadcasting.