Audio, in a feed watched on mute

The constraint

Captions are the video, not a subtitle track

A clip whose meaning lives in the audio has already failed for most of the people it reaches. That reframes the whole craft: timing, weight and line breaks in the captions are the entire visual design of a podcast clip, and treating them as an accessibility afterthought is why most of them look identical and few perform.

The selection problem

The best moment in the room is usually the worst clip

A clip has to make complete sense to somebody with no idea what came before it, which is a much stricter test than being interesting. The moment that landed in the recording usually landed because of twenty minutes of context, and that context is not coming with it. What survives extraction is disagreement, a specific number, or a strong claim stated plainly.

• Pull the complete exchange. Half a question and its answer is the most common broken clip.
• Cut before the setup. The usable clip starts after the speaker began the point.
• Make speaker changes visible. Two voices with no visual distinction is unfollowable on mute.
• Something has to move. A static logo with text over it is an audiogram nobody asked for.

The maths

Three good clips per episode, not ten

A forty-minute conversation usually contains three or four genuinely self-contained moments. Ten is a sign the bar dropped somewhere around number five, and a feed of mediocre clips trains an audience to scroll past your name.

The honest bit

Clips grow the clip account more than the show

Short-form reliably grows the account posting the clips and grows the podcast modestly. The conversion from a short-form viewer into a subscriber who commits to forty minutes is real and much smaller than the view counts suggest, and planning around the larger number is how podcast teams end up disappointed by a channel that is working.