calmer clips: a gentle zoom, captions below your face

the short version

the shorts underway cuts from your episodes now look calmer. the frame moves far less: a gentle push in only where a new thought starts, at most once every seven seconds, eased in over less than half a second instead of snapping. the captions sit lower and smaller, under your chin rather than across your face, and never more than two lines. and the clip finder now prefers the stretches you said cleanly, and takes out words you said twice even when the repeat falls across two lines.

nothing about what a clip is has changed. it is still only your words, in the order you said them, an advert that sends people to the whole episode. it just looks and sounds less like a machine made it.

why it changed

rich, who builds underway, watched the two clips cut from his episode 58, the one about twelve signups and almost nobody getting in. his verdict was short: hard to watch. not the story, the picture. the frame zoomed in, out, in, out on his face on nearly every line, because the old rule was a zoom on every sentence, and a sentence in a real transcript is often two or three seconds long. on a face, that reads as jitter.

the captions were the second problem. they sat a little over halfway down the frame and they were big, which on a talking head filmed at a desk is right across the chin and mouth. the thing a clip most needs you to see is your face, and the words were on it.

the third was the words themselves. the first clip had a few little stumbles left in it: a phrase begun twice, a word repeated across the join between two lines, a run that trailed off into something, something. each one is small. together they make thirty seconds feel long.

what a clip does now

• the frame starts full and stays there for at least the first seven seconds. after that it may push in slightly, to 1.08, or back out, only where one thought ends and the next begins, and only when the next line is long enough to settle on. two short lines in a row never bounce in and out.
• every change eases over four tenths of a second, so it reads as a camera drifting closer, not a cut.
• captions sit in the lower third: the block ends just above the bottom fifth of the frame, clear of the platforms' own buttons and captions, and two lines reach no higher than about seventy percent of the way down. the type is smaller, still easy to read on a phone. the word being said is still lit in your accent colour.
• a long caption is split into two shorter ones rather than wrapping to a third line. no word is dropped and none moves.
• words said twice come out wherever they fall, including across the join between two lines and the ones that only show once an um or a like between them is gone: 'cause, 'cause, or i, like, i.
• when the finder reads your episode it is told which lines stumbled, and it is asked to prefer a run you said cleanly, or to start after the stumble or end before it. a clip full of restarts ranks lower in the list. it is never removed for it.

what did not change

your episode is untouched. the calmer zoom, the lower captions and the tidier words are the clip's, not the episode's. the episode still plays every line as you said it, in order, with its own captions where they always were, which sit low already.

and the clip rules hold: only your words, nothing reworded, a line you struck from the episode never comes back, and the line on screen at the start is copied exactly as you said it. tidying takes out a repeat or a restart; it never stitches words into a sentence you did not say.

clips you already have

clips made before this keep the old look; nothing is redrawn behind your back, because you may already have posted them. to get the new look on an episode's clips, open the episode, go to share and press find clips again. it reads the episode afresh, picks again with the new preference for clean runs, and makes a new set. the old set is replaced on the page; anything already posted stays as it was posted.

the thinking

the punch-in on every line is the grammar of a certain kind of short: the creator who wants to look edited. it works when someone is performing to camera at speed. a founder talking honestly about a bad day is not that, and the same trick makes them look nervous. the brief for a clip was always the same as the episode's: it should look like a well made thing, and the person in it should be the subject, not the edit.

find clips is on the share tab of every finished episode, on the web and the phone. start a series.