Almost everything written about video writing is about openings, and
almost all lost watch time is in the middle. That is not a coincidence: an opening has
an obvious job and an obvious failure. A middle has neither, so it
accumulates. This is a book about the other nineteen minutes.
Rule 1
Fix the order before the words
A well-phrased sentence in the wrong place is still in the wrong place.
When a script is not working, the instinct is to start rewriting lines. It feels productive and it almost never helps, because the common faults are structural and line-editing cannot reach them. A beautiful sentence in the wrong slot is still in the wrong slot, and polishing it makes it harder to move later, because now you are attached to it.
Run the checks in this order, and stop at the first one that fails. Does the video deliver what its opening promised? Is the best material buried somewhere in the middle because that is where it occurred to you? Is anything said twice? Does the emotional weight sit where the structure implies it should? Only when all four pass does phrasing matter.
That last point is worth sitting with. If your opening promises a method and the body spends most of its length on how you felt, the weight is backwards, and no amount of rewriting either part fixes it. One of them has to change what it is.
Run this on your draft
Read only your first line and your last line, with everything between them covered. If they are not about the same thing, you have a structural problem, and it is the only problem worth working on today.
Rule 2
Specificity is the engine
A concrete detail implies the story that produced it. A category implies nothing.
Most flat writing is not badly phrased, it is abstract. Abstractions are closed: nothing follows from them, so the viewer has nowhere to go. A specific fact is open, because a specific implies the story behind it and the mind reaches for that story automatically.
This is why numbers, dates and named objects outperform emotional statements even when the emotion is the subject. "I was afraid of judgment" is a category. "I deleted the video twice before I posted it" is the same fear with a mechanism attached, and it is the mechanism people watch.
The tell is a small set of words that almost always compress something better: partnered, collaborated, reached out, worked together, made it happen, went viral. When you catch yourself reaching for one, the fix is never a better synonym. It is the detail you skipped.
Worked rewrites
What people write
I reached out to a few brands and eventually one of them said yes.
The same material, working
I emailed forty-one brands. One replied, six weeks later, and it was the one I almost did not bother with.
Same event. The right-hand version has a number, a delay and a reversal, and every one of those opens something. "A few" and "eventually" close it.
What people write
The shoot was a disaster and we nearly did not get it done.
The same material, working
We lost the light at 4:40. The last shot is the fourth take, and you can hear the farmer shouting at us in it.
Disaster is a label on the story. The specifics ARE the story, and they are funnier and more credible than the word disaster.
What people write
This gear completely changed the way I work.
The same material, working
I stopped carrying the tripod. My shot count per day went from about nine to about twenty.
A claim you assert versus a claim you demonstrate. The second one cannot really be argued with, which is what makes it land.
Why it bites harder in long form
Over twenty minutes the cost compounds. A short can survive on one abstraction because it is over before the viewer notices. A long-form script made of them is a list of claims, and there is nothing in it to hold on to.
Run this on your draft
Highlight every abstraction in your script. For each, ask what actually happened. If you cannot answer, that line is standing in for material you do not have yet, and it should be a note to go and get it rather than a sentence.
Rule 3
Write the mechanism, not the summary
Summary is what a script fills with when the writer is short of material.
A summary tells the viewer that something happened. A mechanism shows them how, in sequence, with the specifics intact. "Which is how I got my first client" is the label on the story. Who emailed whom, what the message said, and what came back IS the story.
Summary is seductive because it is fast to write and it covers ground. That is exactly the problem: covering ground is not the job. One concrete scene beats any amount of ground covered, and a script that summarises three things well would almost always be better as a script that shows one of them properly.
There is a reliable signal for this. If a section of your script could be replaced by a single sentence without losing any information, it is summary, and it is taking up minutes that a scene could have used.
Worked rewrites
What people write
So I spent the next few months learning colour grading and it made a huge difference to my videos.
The same material, working
The first LUT I ever made turned skin green. I spent a week finding out it was one setting, and the setting was not the one anybody mentions.
The left covers months and says nothing. The right covers a week and leaves you wanting the setting. Less ground, more video.
What people write
We had a few problems with permissions but got there in the end.
The same material, working
The security guard asked what we were filming. I said a school project. I am thirty-four.
One scene, three lines, and it does everything the summary was trying to do while also being funny.
Why it bites harder in long form
This is the single biggest difference between a short and a long-form script. Short form has no room for summary, so it never learns the habit. Long form has nothing but room, and fills it.
Rule 4
Write for the ear, not the page
The viewer hears it once, cannot re-read it, and does not rewind. They leave.
A script is spoken. This sounds obvious and is routinely ignored, because we write with our eyes and a sentence that reads fine can be unfollowable out loud. When a listener loses the thread of a sentence they do not scroll back, they go somewhere else.
The rules that follow are all consequences of one constraint: meaning must arrive in the order the words do. A reader can hold a suspended clause because their eye can jump ahead. A listener cannot.
One idea per sentence. Put the subject at the front. Avoid clauses that hold the meaning hostage until the end. Say numbers the way a person says them, rounded, unless the exact figure is the point, in which case give it alone. And use contractions, because "do not" and "it is" said out loud sound like a press release.
Worked rewrites
What people write
After spending nearly three months failing at this and very nearly giving up on the whole idea, the thing that finally worked was almost embarrassingly simple.
The same material, working
The thing that finally worked was embarrassingly simple. It took me three months to find it.
The left suspends its meaning for twenty-two words. The right delivers the point, then the cost. Same content, and the second one can be said in one breath.
What people write
Revenue for the quarter came in at £19,847.
The same material, working
We made just under twenty grand. Nineteen eight four seven, if you want the exact number.
Nobody hears a six-digit figure. Round it for the ear, and give the exact one separately if it matters.
Run this on your draft
Read your script out loud, standing up, at the pace you actually talk. Every place you run out of breath, stumble, or have to re-read is a place the viewer would have lost it. Mark them and split them.
Rule 5
Do not resolve in one breath
Setup and payoff in the same sentence means nothing is left hanging.
This is the most common line-level fault in scripts and the easiest to fix, because the fix is punctuation rather than words.
A sentence that contains both its tension and its release is closed. The viewer receives it as information, and information does not pull anyone forward. Split it and the same words become open: the first half creates a question, the pause holds it, the second half pays it.
It applies anywhere a reveal exists, not only in your opening. Every time you catch an "until", a "but then", or a "which turned out to be" in the middle of a sentence, you are probably looking at a full stop that has not happened yet.
Worked rewrites
What people write
I thought the shoot was ruined until I checked the second card.
The same material, working
I thought the shoot was ruined. Then I checked the second card.
Identical words, one full stop. The left is a fact; the right is a beat with a gap in the middle of it.
What people write
It cost four hundred pounds, which turned out to be the best money I ever spent.
The same material, working
It cost four hundred pounds. I have never been so glad to waste money.
Splitting also let the second half become a sentence with a shape of its own, rather than a subordinate clause hanging off a price.
Run this on your draft
Search your script for "until", "but then", "which turned out", "only to find". Each one is a candidate for a full stop.
Rule 6
The middle is where long-form dies
Openings and endings have obvious jobs. The middle has none, so it sags.
Almost all advice about video writing is about openings, and almost all lost watch time is in the middle. That is not a coincidence: an opening has a clear obligation and a clear failure mode, so people work on it. A middle has neither, so it accumulates.
Three things keep a middle alive. Each segment opens its own small gap: not a new hook for the whole video, just a reason this section is worth its minutes. A segment that begins by announcing what it is about to cover has already spent its only tension.
Transitions are exits. Every boundary between segments is a place where the viewer is momentarily finished, and a finished viewer is a free viewer. Write the last line of a section so it leans into the next one instead of wrapping up neatly. Resolution is what you owe them at the end, not every four minutes.
And never make the same point twice. Length invites restatement, because there is always room and it feels like emphasis. It reads as the writer forgetting. The viewer already has it.
Worked rewrites
What people write
So that is the lighting sorted. Next, let us talk about audio.
The same material, working
So that is the lighting sorted. Which created a problem I did not notice until I got home.
The left closes a section and announces the next one, giving the viewer a clean place to leave. The right opens a gap across the boundary, so leaving means not finding out.
What people write
In this next section I am going to show you three ways to fix this.
The same material, working
There are three fixes. Two of them made it worse.
Announcing a section is the same failure as announcing a video, one level down. State something with a hole in it instead.
Why it bites harder in long form
This chapter has no equivalent in short form, which is why so much short-form-trained advice does not transfer. Sixty seconds has no middle.
Run this on your draft
Open your retention graph and find where it slopes hardest. Now find that timestamp in your script. It is almost always either a transition that resolved cleanly or a point you had already made.
Rule 7
The promise works at two scales
The title makes one. Every beat makes a smaller one.
Your title and thumbnail made a promise, and the opening has to visibly start paying it. That much is well understood. What is less understood is that the same contract runs at the level of every individual beat.
A beat makes a small promise in its first line and owes payment by its last. When a beat opens on one thing and closes on another, it has drifted. That is usually a structure problem rather than a writing one, and it is worth naming as such instead of trying to smooth the seam with better sentences.
The end of the video is where the top of it gets collected. If your opening planted a stake, a question or an image, that is what the close is for. An ending that introduces a new idea leaves the opening uncashed, and the video feels like it stopped rather than finished.
Run this on your draft
Write down, in one sentence, the promise your title makes. Then read the first line of every beat in order. Any beat whose first line has nothing to do with that sentence is either in the wrong video or in the wrong place in this one.
Rule 8
Metaphors, and where they go
Literal first, image second, and almost never in the opening.
A metaphor compresses something the viewer is already holding. That is its whole function, and it is why order matters: lead with the image and you have handed them a picture attached to nothing, which they then have to decode while the video moves on without them.
Position follows from that. At a turn, a metaphor recasts what just happened, which is useful. At the close, it generalises past you to them, which is powerful. In the opening it is a lateral move at the exact moment the viewer needs forward motion, which is why so many scripts open with one and lose people anyway.
Spend half and save half. An image with two parts can open at the turn and collect at the close. Spend it all at once and your ending has nothing to call back to. And one per video is usually right, because two compete and the viewer remembers neither.
Rule 9
Do not inflate
Your audience audits you, out loud, in the comments.
One inflated claim makes a viewer re-examine everything else in the video, including the true parts. This is the cheapest possible thing to get right and the most expensive to get wrong, because the damage is not to the claim, it is to you.
Cut overclaims: the sole reason, the best way, nobody talks about this, starting from scratch when you did not. Cut self-conferred authority: as a filmmaker, as an expert in. Showing beats claiming, and the claim actively undercuts the evidence sitting next to it, because a person with the evidence does not need the label.
Watch for the figure a knowledgeable viewer would dispute. Often the fact is fine and one attached word is what is contestable: the number is right but "ever" or "only" or "sole" is not. Drop the word, not the sentence.
And watch for drift. People describe the honest version of a story in conversation and then write a safer one, because writing is done to be seen. If you told a friend this story and it came out differently, the version you told your friend is nearly always the better script. That is not a rule about honesty, it is a rule about quality: unguarded language is more specific, and specificity is the engine.
Worked rewrites
What people write
As a professional filmmaker, I can tell you this is the single best camera under a thousand.
The same material, working
I have shot four paid jobs on this. Two of them I would not have taken with the old one.
The left claims standing and a superlative, both of which invite argument. The right demonstrates standing and makes a claim nobody can dispute because it is about you.
What people write
This one change doubled my views overnight.
The same material, working
The next video did about twice the views. I have changed three things, so I cannot prove it was this one.
Admitting the confound costs you nothing and buys you everything. The viewer was going to think it anyway; saying it first makes you the reliable narrator.
Where this sits
The Hook Book covers the first fifteen seconds:
the five patterns and what to say immediately after each one. This book is everything
after that. If you want the openings formulas rather than the craft, the
hook library has fifty-eight of them, and the free
hooks module walks through writing one from scratch.