Best way I would know how to do the video summarization (or at least what I would attack first) would be with a combo of whisper ai and then a solid LLM. The first would be to transcribe the video, then feed the output into the LLM with a prebuilt prompt with some summarizing instructions/specifics. Would either need something like Langchain (for chaining Ai operations), or you could use ChatGPT to write the code to bridge them together with just javascript or something.