I've never upgraded to a real video editor so I wind up relying on ffmpeg after using one of free ones (if I make any tweaks at all instead of just cutting to a

I've never upgraded to a real video editor so I wind up relying on ffmpeg after using one of free ones (if I make any tweaks at all instead of just cutting to a segment). It looks like most projects on GitHub are interested in reading markers rather than generating the FCPXML. Interesting point on your demo video re: importing from YouTube and enabling hardware acceleration with h_264videotoolbox for the conversion. Now you just have to find a good search term, pull down the most viewed videos and use YouTube's "most replayed" graph with AI-powered cleanup on where to cut to create an automatic best-of aggregator!