[fareesh]

What's the high level approach required to build something like this yourself?

Does it involve relying on speech to text with timestamps and then a series of cuts based on that?