This repository hosts an app developed using Whisper and Langchain that allows the creation of a Q&A assistant and video summarization. The model's maximum context length is 4097 tokens (gpt-3.5-turbo).
The App can be run locally but requires an OPENAI_API_KEY in the .env file. Feel free to ⭐ and clone this repo 😉
I recommend installing the modules in the following order. The ffmpeg module is required for the proper functioning of the application. You can install it using Conda as follows:
poety install
The app can be used running streamlit run youtube_llm_app/app.py in the terminal. There are 2 options on the sidebar, Q&A or Summarize. I recommend using videos no longer than 5 min of speech due to the model tokens' limitations.
The first option allows a Q&A assistant to ask questions about the video.
The second option allows us to get a summary of the video.
