Chop your videos up the easy way using Whisper to find the word breaks. works best with compute power from google colab or similar T4+ levels of compute but might work from some home setups just via CPU, haven't tested yet. requirements.txt coming as soon as someone complains about it.
There is a neat trick where-by the output timestamp text file can be used with a LLM like ChatGPT, Claude, Gemini etc. to seperate the clips into segments that are clustered by subject. As soon as I can find that bit of email exchange I'll put it up here but most of you should be able to figure out how on your own with a little trial and error.