Art of Automating Short-form Sports Video

At this very moment, artificial intelligence (AI) is redefining sports by extracting insights through computer vision (CV) and machine learning (ML): machines are autonomously racing 400+ HP cars on the Monteblanco track, another machine is advising a coach on the best strategy for an upcoming game based on millions of data points, while another is analyzing a golf swing to identify a flaw in its motion.

Machine learning is revolutionizing innovation in analytics, prediction optimization, automation, simulation and decision-making. Computer vision is automatically generating game highlights, live production streams, detection of play types and team formations on the field, and analysis of player and ball movements.

However, CV and ML are not without limitations. Tracking a soccer ball and athlete movements on the field is possible - and might be accurate - but is it a "great play,” or the moment an unsung hero makes a crucial stop to clinch the victory? When a goal is scored, is its impact on the game clear? Not all scoring plays are created equal; a score that puts a team up four goals in the first half is different from a goal that ties the game in the final seconds.

For me, the most fascinating aspect of sports video lies in capturing compelling moments and igniting the emotional connection fans have to them - it’s incredibly exciting to be building technology that can support this. But to deliver what audiences truly crave in a video experience - accuracy, context and emotion - human input is required. While a machine cannot understand higher-level concepts like rivalries and their potential impact on a game, humans can. Game video highlights delivered without expert human commentary feel mechanical because they’re devoid of the fun and the energy that is intrinsic to sports.

At WePlayed Sports, we believe the human element is an irreplaceable part of creating content that inspires engagement with the fans and athletes. That’s why we built an AI-driven sports video platform that leverages CV and ML to create a rich library of short-form videos we call #moments. This library is searchable with advanced sport-specific capabilities, and it is the foundation from which #moments, playlists and recaps can be created for games, teams and athletes.

Our A.I. does the heavy lifting, locating the plays people care about. It inspects what's happening on the screen (game clock, scoreboard and other visual clues) to create the optimal short-form video #moments. It then provides a variety of tools to help people find #moments, add custom graphical annotations, tweak timing, and, most importantly, provide the emotional and expert context for why the play is worth watching. #Moments can be automatically or manually placed in playlists and profiles. But, the creation of short-form video highlights is just one use case our product supports.

Today, we support 12+ different sports and the list will continue to grow in 2021. We are also working on live-game capture and a clipping feature to pair with real-time game play-by-play feeds.

Our platform is built to serve everyone from coaches and school athletic department administrators to fans, video bloggers and former athletes. For all of these people, the goal is simple: to discover and share exciting and inspiring sports stories, told through video.

This is why the team at WePlayed Sports believes that Every #Moment Counts.