Inspiration 💡

Sports games can be long and uneventful at times coming from our personal experience watching playbacks of boring Super Bowl games and World Cup matches. The sports entertainment experience can be greatly streamlined and enhanced if we present users with the highlights and key moments of a sports game which is what SportsTLDR does with the use of genAI.

What it does 🏈

SportsTLDR allows the user to input a YouTube link to a sports video and then processes the video to produce a text transcript detailing the key moments of the game. It also includes the time stamp for indicating when a particular key moment was detected for ease of access so the user can simply rewind to that time stamp and watch the most exciting goal or touchdown, whatever highlight it may be.

How we built it 👨‍💻

We built SportsTLDR using a React frontend and a Python and Flask backend. We used the YoutubeTranscriptAPI to extract the full transcript of the user's inputted sports game video. Further, we trained the T5 Model for completing the NLP task of returning the highlights of a sports game video based on the transcript.

Challenges we ran into 🔧

Some challenges we ran into include choosing the AI model for performing NLP tasks, using the API to obtain the video transcript off of YouTube, and initially connecting the frontend to the backend code for our application.

Accomplishments that we're proud of 🚀

We are proud of SportsTLDR's ability to accurately return a description of the highlights of any and all sports games when simply given the link to the video.

What we learned ✍️

We learned more about the ability of generative AI to perform a variety of tasks, which in our case involved video editing and extracting desired elements from a video as well as leveraging AI models to perform NLP tasks.

What's next for SportsTLDR 📺

In the future, we seek to add features for SportsTLDR to generate a video version of the highlights it extracts, and for the user to conveniently see this key moments video in our application. We also would like to use this application together with different methods of streaming sports games, not just on YouTube playbacks, such as Hulu, Paramount, FuboTV, ESPN+, and more.

Share this project:

Updates