The middle of a script problem

ViewMade renders every short at 1080x1920 with captions burned into the frame, and as of 22 August 2026 the product ships 7 caption styles for that output. The format is fixed. What varies between a short people finish and one they abandon is the script, and the part that breaks most often is the middle. This article explains why the middle of a script fails, gives a beat structure that holds it together, and tells you what to do on your next draft.

The short answer

The middle of a script fails because it has no job. The opening has one: earn the next ten seconds. The ending has one: pay off the promise the opening made. The middle is where writers stack facts without order, and viewers read that stacking as "nothing is happening yet" and leave. The fix is to give every middle beat one of three jobs: raise a question, answer the previous question while raising a new one, or complicate what the viewer assumed. If a sentence in your middle does none of these, cut it or move it.

Why the middle loses viewers

A short is a contract. The first line states a claim or a question, and the viewer stays only as long as they believe the answer is coming. The opening spends that trust. The ending repays it. The middle is where trust is spent fastest, because the viewer has nothing new to evaluate: they are waiting.

Three failure patterns cover most broken middles.

First, the list. The writer has five facts and delivers them as five facts. Each fact is fine on its own, but none changes the meaning of the one before it. The viewer learns nothing they could not have guessed from the title, and the title already did its job.

Second, the detour. The writer adds context that was interesting during research. Background on a person, a date, an institution. It reads as padding because it does not move the claim forward, and in a vertical short the viewer's thumb is always one swipe from ending the contract.

Third, the plateau. The script peaks in the first third. The most striking image, the most surprising fact, lands early, and everything after it is quieter. Viewers who came for the peak have already had it.

All three have the same root: beats were ordered by when the writer learned them, not by what each one does to the viewer's next question.

A beat structure for the middle

Treat the middle as a chain of questions. Each beat answers the question the previous beat raised and raises exactly one new one. The viewer never has a moment where no question is open, and never has two questions competing for attention.

A workable structure for a 60 to 90 second documentary short:

Beat	Job	What goes here	Common failure
Hook	State the claim or question	The most concrete, checkable fact you have	Starting with context instead of the claim
Turn 1	Complicate the assumption	Evidence that the obvious answer is wrong	A fact that merely adds detail
Turn 2	Raise the real question	The detail that makes the viewer ask "then how?"	Explaining too early
Turn 3	Answer, with a cost	The resolution, and what it cost someone	A resolution with no consequence
Close	Land the last line	The sentence the viewer repeats	Restating the hook

Two rules keep the chain tight. One question per beat: if a beat answers two things, split it. No orphan facts: if a fact does not raise or answer a question, it belongs in a longer video, not this one.

Notice what is not in the table: percentages, retention curves, ideal durations. We have not measured those, and a structure that claims a number it did not measure is worse than one that admits the limit.

What to do on your next draft

Take your current script and mark each sentence with one of four labels: H (hook), Q (raises a question), A (answers one), or N (neither). Then do three passes.

Pass one: delete every N. Most first drafts of a middle are half N sentences. This is the fastest quality gain available, and it costs nothing.

Pass two: check the alternation. You want a pattern close to Q, A, Q, A through the middle. Two Qs in a row means you opened a question and walked away from it. Two As in a row means you are lecturing. Reorder or merge until the alternation holds.

Pass three: read only the first sentence of each beat, out loud, in order. That sequence should tell the whole story on its own. If it does not, the middle is carrying weight the beats cannot see, and you have found the exact sentences to restructure.

If you work from a topic rather than a draft, the same structure applies at generation time. ViewMade builds its scripts as a question chain for exactly this reason: research produces far more facts than a short can hold, so the script stage is a filtering stage, and the filter is "does this beat raise or answer a question." The output is a finished vertical documentary with sourced archive footage, word-for-word captions, and a media credit file naming the source of every clip, measured at 22 August 2026.

What this does not tell you

This article describes structure, not measured outcomes. We have not measured how beat order affects retention, watch time, or completion rate on any platform, and we will not attach a percentage to a structure we have not tested. The claim here is narrower: the three failure patterns above account for the broken middles we see in practice, and the question-chain structure is a usable repair.

The advice also has limits. It is written for vertical documentary shorts in the 60 to 90 second range, on topics where a claim can be stated in one line: a disaster, a heist, an unsolved case, a historical turn. It transfers poorly to entertainment formats where the appeal is a personality, a reaction, or a joke cadence, because those scripts run on timing rather than open questions. And it assumes the viewer arrived from a title that made a specific promise. If your title is vague, no middle structure will save the contract, because the contract was never written.

<!-- faq -->

Frequently asked questions

How long should the middle be compared to the hook and the close? In a 60 to 90 second short, the middle is most of the runtime, typically three beats against one hook and one close. The useful constraint is not a ratio but a count: three turns is enough for a single claim, and a fourth turn usually means you have two videos. If your topic needs five turns, split the topic.

Should I put my best fact in the hook or hold it for the end? Put a strong, concrete, checkable fact in the hook, but not necessarily the single strongest one. The hook must state the claim; the strongest fact often works better as Turn 3, where it answers the question the middle has built. If your best fact can serve as the claim itself, lead with it and build the middle around what it cost.

How do I know if a sentence is an N sentence? Ask what the viewer can do with it. If the sentence adds a detail that does not change what they expect next, it is an N. Dates, names, and institutional background are N sentences more often than writers expect. Keep one only if a later beat depends on it.

Does this structure work for list-style shorts? Partly. A list is a chain of As with no Qs, which is why list shorts live or die on the strength of individual items rather than structure. If you want the question-chain effect, order the list by escalation, so each item makes the next one harder to believe, and close on the item that recontextualizes the first.