How long should a short-form clip be?

There is no correct length, and the question is worth asking anyway. What follows is the reasoning we use when cutting, and the two places where the answer is actually decided.

Length is an output, not a setting

A clip is as long as its idea. The unit that travels is a complete thought: a question and its answer, a claim and the evidence for it, a story with a turn in it. Cut before the turn and the clip is a tease nobody rewards; run past the resolution and you are asking for attention you have already spent.

This is why "make it 30 seconds" is the wrong instruction to give an editor, and why an AI that picks a fixed window around a keyword produces clips that feel almost right and perform badly. The moment has a natural start (the setup) and a natural end (the beat after the payoff). Find those two points and the duration is whatever it is.

The two numbers that actually matter

Watch-through rate. Every short-form feed is optimising for how much of a video people finish. A 25-second clip watched to the end beats a 55-second clip abandoned at 30, and it beats it on the exact metric the ranking uses. Length is therefore a bet: every extra second has to carry its own weight, because it raises the number of people who leave before the end.

Time-to-hook. Whatever the total, the first two seconds decide the rest. If the setup takes eight seconds to reach the interesting part, the clip is not too long, it is badly started: move the payoff line to the front as a hook, then let the setup follow. This is the single edit that changes results most often, and it costs nothing but the decision to make it.

Practical ranges, and what they are for

• Under 15 seconds. One line, one reaction, one number. Very high completion, very little room to say anything. Good for a punchline, useless for an argument.
• 15 to 45 seconds. Where most clips that work land. Long enough for a question and a real answer, short enough that the whole thing gets watched.
• 45 to 90 seconds. For a story or a demonstration that genuinely needs the time. Every second of it has to earn its place; this is the range where a lazy cut shows.
• Past 90 seconds. Occasionally right, usually a sign that the clip is really two clips. Try splitting it and see whether both halves stand up.

Platform maximums move around, and they have moved several times while this product has existed. Check the limit where you are posting rather than trusting a number you read in a blog post, including this one.

Cutting to length without cutting badly

The temptation, once a clip runs long, is to trim the top and the tail until it fits. That usually removes the setup and the payoff and keeps the middle, which is the worst possible trade.

Cut from the inside instead. Spoken language is full of restarts, filler and sentences that arrive at the same point twice. Removing those tightens the clip without touching its shape, and it is the reason editing by transcript is faster than editing by waveform: you can see the redundancy as text, strike it, and let the cut ripple through.

If the clip still runs long after that, it is not too long. It is two ideas.