The fastest way past a blank script page is to pick a structure that matches the job the video is doing, then fill it. These are eight structures that reliably work for talking-head video — quick tip, problem to solution, three-step how-to, answer a comment, product demo, personal story, opinion, and the long-form opening — each with a performable skeleton and a filled example you can adapt.
They're organised by job rather than by platform, because the same structure becomes a TikTok, a Reel, a Short or a LinkedIn video just by changing its target duration and its close. All eight ship as starter templates in TellyPrompter, pre-filled so you can read one, replace the words with yours, and take the take.
Why structure beats length
The question people ask is "how long should the video be", but length is downstream of a better question: what job is this video doing for the viewer? Teaching one thing, fixing a pain, walking a process, answering a person, showing a product, telling a story, taking a position, or opening something longer — each job has a shape that works, and the shape tells you what to write next when you stall.
A structure is also what makes a format repeatable. One good "quick tip" is a video; a quick-tip structure you can refill weekly is a series, and series are what audiences actually subscribe to.
The durations below are targets to write towards, not promises about performance — appropriate length depends on the content, and your own measured read time is the only number that decides.
1. Quick tip
The job: teach one immediately useful thing. Target: 15–60 seconds.
The skeleton: direct promise → the tip → a miniature example → takeaway or next action.
Filled: "Your prompter font is too small — here's the fix in ten seconds. Set the size from where you'll actually stand, not from the desk. I read a line from my mark; if I lean in, it goes bigger. Try it before your next take and watch your eye line settle."
The discipline is the single tip. If a second one creeps in, it's next week's video.
2. Problem to solution
The job: help a viewer recognise and fix a pain. Target: 30–60 seconds.
The skeleton: relatable problem → why it happens → the solution → the result → next step.
Filled: "You sound flat on camera and you can't work out why. It's usually not you — it's the scroll. A fixed-speed prompter makes you perform to a metronome, and the flatness is your attention split in two. Let the script follow your voice instead, and the take relaxes. Next video: the settings that make it work."
The relatable problem must genuinely be relatable — the viewer's own words for it, not yours.
3. Three-step how-to
The job: teach a repeatable process. Target: 45 seconds–3 minutes.
The skeleton: the outcome first → step one → step two → step three → recap.
Filled: "By the end of this you'll have a phone teleprompter rig for under thirty pounds. Step one: the hood — here's the one I use and why. Step two: mirror mode, set before the phone goes in. Step three: font size from your mark. That's it — hood, mirror, size — and you're reading straight down the lens."
Outcome first is what earns the steps. Three is a shape, not a law — but past five, it's a course, not a video.
4. Answer a comment
The job: turn audience demand into content. Target: 15–60 seconds.
The skeleton: show or paraphrase the question → direct answer → the nuance or an example → invite the next question.
Filled: "Sam asks: does a teleprompter work if you wear glasses? Yes — with two adjustments. Raise your key light and tilt it down, and use a dark background so there's less screen to reflect. Here's what mine looks like on camera. What should I answer next?"
Answer first, nuance second. Making them wait for the answer is the one way to waste this format — and the platforms treat comment replies as a native workflow, so the demand is already measured.
5. Product demo
The job: show how something works. Target: 30–90 seconds.
The skeleton: the problem or use case → the thing in action → the benefit made visible → an honest limitation or who it's for → next action.
Filled: "Recording alone means nobody's operating the prompter. Watch what happens when I pause mid-sentence — the script waits. When I ad-lib — it waits again, then picks me up. That's a take with no operator. It does need a browser with speech recognition, so not Firefox. Link below if your rig needs this."
The honest limitation isn't a courtesy, it's what makes the rest believable. If you can't name one, you haven't used the product enough to demo it. (If the demo is sponsored, that's a different structure with disclosure obligations — deliberately not this one, and covered properly in a follow-up piece.)
6. Personal story with a lesson
The job: build connection without rambling. Target: 60 seconds–3 minutes.
The skeleton: a moment of tension → what changed → one specific detail → the lesson → the viewer's takeaway.
Filled: "Take eleven, and I was getting worse, not better. I nearly scrapped the whole video. Then I noticed something in the playback: my eyes, flicking left every four seconds — to the script beside the lens. Moved the camera two metres back, retook it once, done. The lesson wasn't practice; it was setup. If your takes get worse as you go, stop performing and check the rig."
The specific detail — the eyes, the four seconds — is the story. Without it you have a fortune cookie.
7. Opinion or myth correction
The job: state a useful point of view. Target: 30 seconds–2 minutes.
The skeleton: the common belief → your position, plainly → two reasons or examples → a fair caveat → question or close.
Filled: "Everyone says memorise your script so you look natural. I think that advice makes most people worse. Half-memorised is the worst state — you recite from memory and check the prompter, and the checking is what looks shifty. And memorising burns the hours that should go into the writing. Fair's fair: for a 20-second scripted gag, memorising works. For everything else, trust the prompter. Disagree? Tell me why."
The caveat is load-bearing: a position with no conceded ground reads as a pose.
8. The YouTube opening 30 seconds
The job: start long-form without throat-clearing. Target: a 30-second opening; the body is its own script.
The skeleton: meet the title's promise immediately → show the payoff or stakes → preview the route through the video → begin section one.
Filled: "This is a complete beam-splitter setup for one person — by the end you'll have built it and taken a test read. Here's the finished rig, and here's the take it produced. We'll do glass, mount, mirror, light, then the two-minute check. First, the glass, because it decides everything else."
No greeting, no channel biography, no "before we start". The first 30 seconds are measured as their own unit, so write them as one — and put the best material forward rather than saving it.
Adapting one structure across platforms
Resist making four versions of the taxonomy. One structure travels: take the quick tip above, and it's a 15-second vertical clip with a "follow for more" close, a 45-second YouTube Short with a related-video close (a spoken one — links aren't clickable in a Short's description), or a LinkedIn post with a professional-context close. What changed was duration and the final beat; the structure didn't.
Save what works as a series
The last step is the one that compounds: when a structure lands with your audience, keep it. Save the script as a reusable format — in TellyPrompter, that's a script in your library with a tag like "series: quick tips" — and refill it rather than reinventing it. Comparing like with like is also the only honest way to learn what your audience wants: eight videos in one repeatable format teach you more than eight one-offs ever will.
Structure is scaffolding, and scaffolding comes down. Once one of these shapes has carried you through a first take, you'll bend it, and the bent version — the one that fits how you actually talk — is the real template. These eight are just the fastest way to get there from a blank page. For the writing itself — the read-aloud pass that makes any of these performable — see how to write a script you can actually say out loud.