Automating Cricket Narration Using LLMs and Multi-Modal Analysis

Sports commentary enhances audience engagement by delivering analysis, emotion, and context. This paper proposes a novel system for automated cricket commentary generation using a Large Language Model (LLM) and multi-modal data extraction. The framework processes match videos to classify key events such as batter shots, fielding actions, and player sentiments, integrating scoreboard data and historical statistics for context-aware, delivery-specific narratives. A machine learning-based classifier detects events, which an LLM transforms into human-like commentary by combining real-time insights, sentiment analysis, and statistical context. Experimental results demonstrate high event classification accuracy and coherent commentary generation, highlighting the system’s potential to automate and enhance sports broadcasting.