News Cap

News Cap - A Fact Checker

Inspiration

More than ever, people are getting their information from social media. Built in "fact checkers" are often one dimensional and don't show the full picture. We wanted to build something that protects free speech and yet promotes a diversity of opinions right where viewers are watching.

What it does

News Cap is an AI-native video player, making valid research-backed information more accessible than ever. Simply upload a YouTube link and our Agentic workflow will research the claims made in the video and cross reference it with real, credible sources. If the user wants to dive even deeper into the topics mentioned in the video, they can easily ask follow up questions.

How we built it

First, we grab the YouTube link and download the video. Then we host it as a public downloadable file and pass that link to Twelve Labs. This embeds the video and allows us to analyze it for any instance where the video is trying to convey facts or convince the audience. Detailed summaries of these clips are then sent to Exa (a Perplexity competitor) to research and find sources on. Then, our app determines the validity of the original claim and returns a concise list of the findings along side the video.

Challenges we ran into

Learning new technologies is always a fun challenge to overcome. One of the largest challenges was determining how to use Twelve Labs to achieve what we were trying to build. With such a wide range of features in their API it was a little daunting trying to figure out how to architect our product.

Accomplishments that we're proud of

This was a very ambitious project, trying to stitch together so many different technologies and build a polished product. 3 miroservices and a front end tested our ability to develop in parallel and ensure that everything would function together in the end.

What we learned

From the beginning we wanted to familiarize ourselves with new frameworks. Some of the cool things we learned were how to use Twelve Labs, Exa, and Cedar-OS. We also enjoyed learning some more advanced design patterns with FastAPI!