Misinformation is widespread nowadays and it can be hard to distinguish fact from fiction, especially in important moments such as presidential debates or speeches where the fate of a country may lie in the hands of the people who are watching.
We created a web application that transcribes text from videos and scrapes the internet for data about what is said in a video. A decision is made by an LLM about what is said based on the evidence that is gathered.
The general shape of the solution allows users to feed us a video which we transcribe using OpenAI's Whisper, an automatic speech recognition (ARS) system, and check for facts or misinformation. The end result displays a list of facts and misinformation gathered from the video.
- Users can paste a YouTube video link into the page to analyze the facts and false claims said in the video.
- OpenAI's Whisper transcibes the video
- GPT 3.5 determines what statements are claims or just opinions / yapping.