Why are you not just using yt-dl to download the subtitles? Using yt-dl you can also fallback in case the subtitles don't exist and instead download the videos, convert them to mp3, and then run them through whisper[1] locally.
You can just paste the URL to gemini.google.com with prompt "Summarize this video."
"I don't have enough information to do that. I'm a large language model, able to communicate in response to a wide range of prompts and questions, but my knowledge about that specific video is limited. Is there anything else I can do to help?"
It'd be great to have a tool to get accurate transcripts. The YouTube subtitles can be easily downloaded, but aren't readable on their own. They often lack even basic punctuation or sentence structure.
It'd be even more great to get good condensed transcripts, or event targeted condensed transcripts of the things I'm interested in in that particular video.