@Gargron hmm. maybe this should be a github issue to get longer-form responses?

My quick thoughts:
remove the "-vsync cfr" and limit duration by time (-t) rather than frame count.
Maybe have a max video size (height/width) for the "original" quality (1080p?).
Can probably get away with a higher (lower quality) -cfr value. Would have to encode some samples to test the balance between -crf and the -maxrate/-bufsize settings.

vp9 would be *nice to have*, but it's too slow to encode...