Long clip cut-down

Cut my long talking video down to a 9:16 short of about 30 seconds. Open on the strongest sentence of the whole clip, tight on my face, then play three to eight stretches in the order that tells one idea, with jump cuts that drop the pauses and word captions under my chin. Footage: I attach one clip over a minute long of one person talking, vertical or horizontal, with its own sound. Transcribe it first. A horizontal clip fills the vertical frame with no bars, moved so the face is centred and the eyes sit on the upper third line. I may also attach B-roll of the things I talk about. Hook and stretches: The hook is the one sentence a stranger would stop for, 0.8 to 3 seconds, played first and tight, with its first word on screen inside the first second. Cut it out of the body so no line is said twice. Then the stretches play, each starting just before a sentence and ending just after a full stop, in the order that builds the idea, not the order I said them. Keep every sentence whole. Cuts: Inside the stretches, cut on sentence and clause starts about every 2.7 to 3.6 seconds, with uneven lengths and no shot under one second. Drop every pause over half a second, but leave 0.1 to 0.25 seconds of air at each cut. Every cut flips the framing between the wide frame and a punch of about 15 percent, with the eyes on the same line. Each shot creeps in 2 percent and never pulls back. Captions: Put the captions under my chin, centred inside the safe zone, never on the face. Use Montserrat 900 capitals at about 80 pixels, white with a black outline and no box. Show one word at a time for a close-up, or two to four words with the spoken word in an accent colour from the footage for a wider shot. The words the idea turns on take the highlight colour. B-roll: When I give B-roll, lay a shot over a phrase only when the words name something it shows, for 1.5 to 3 seconds, never in the hook and on at most half of the short. My voice keeps playing under it. Each B-roll shot pushes in a little, with a soft whoosh at 35 percent as it lands, and the captions move clear of the product while it plays. For a lesson or a sermon, add a short label in the top corner away from my face and a thin progress bar along the top edge. Ending: End on my last sentence, with no end card. If I give a brand line, show it as one sentence in the corner for the last 2.6 seconds. Look and sound: One look on every shot, with contrast and saturation up a little and a soft vignette. My voice plays at full level on every shot. An optional music bed with no vocals sits at about 6 percent, fading in over 0.6 seconds and out over 1.2 seconds.