Aim: Given a video of a tennis match, we aim at automating commentary generation without human intervention.

Motivation

An often-overlooked aspect of a successful video production is the audio.Surprisingly, even if the quality of the video images is poor, as long as the audio is of good quality, the video will still be watchable.A justification of the above fact is the role of good commentary in sports matches.A good commentary heightens the level of para social interaction between the players and the audience,help infer better conclusions about body language of the player and enhance our level of enjoyment. Our project attempts to generate commentary for sports matches automatically without human intervention.We use tennis as a prototype for our demonstration.We rely on sports specific factors such as the continuous video feed,motion of the ball,sound produced when the ball hits the racket,position of player with respect to the court,etc to make useful inferences and generate automated commentary. The project can be extended to generate a full match report capturing important events based on the commentary generated.The approach can be generalized to obtaining useful audio based output for any video input given sufficient information of important video characteristics.