I did something similar piping the output of the youtube-transcript-api python package to openAI's api: https://github.com/DavidZirinsky/tl-dw/