How to script a 90-second sales video you can actually deliver
Writing and delivery are the same problem. The last sentence first, the ninety-second skeleton, writing for the ear, one script per branch, then the prompter near the lens, free takes, and the honest reason the script is yours to write.
The ninety-second sales video is the workhorse of every funnel we have described on this blog: the opener before a choice point, the branch that speaks to one problem, the answer to one objection, the clip that sits in front of a calendar. It is short enough to write in an hour and record in ten minutes, and most people still do not make them, for one reason: the gap between a page of words and a person on camera saying them naturally. This is the guide to closing that gap, writing and delivery together, because they are the same problem.
Write the last sentence first
Before anything else, write the one sentence the viewer should be able to say after watching: “This is for teams like mine who lose leads in the handoff, and the next step is a fifteen-minute call.” If you cannot write that sentence, you do not have a video yet; you have a topic. Everything else is built backward from it.
Use the ninety-second skeleton
Ten seconds: the problem, in the viewer’s words, not yours. Twenty seconds: why it costs more than they think. Thirty seconds: the turn, what is different about your way of doing it. Twenty seconds: proof, one concrete thing that happened. Ten seconds: the ask, which is one action, said once. That is roughly 220 words at a speaking pace. If the draft is 400, you have two videos, and the second one is probably a branch.
Write for the ear, then read it aloud
Short sentences. Contractions. No word you would not say to a customer across a table. Read the draft out loud once and cut anything your mouth stumbled on; the stumble is the script telling you it is written, not spoken. Put the ask in the script as a line you will actually say, because in an interactive video the ask is not a button under the player, it is the moment you turn to the camera and say “answer this one question for me.”
Script each branch as its own clip
A branching video is several short scripts, not one long one. Each branch carries one message for one kind of viewer and ends at the same place. Writing them separately keeps each clip honest about its single job, and makes recording them in sequence painless. The structure tells you what each clip has to say; that is the part people expect to be hard and is not.
Put it on the prompter, eyes near the lens
Reading from a script looks like reading when the text is far from the camera and the eyes track it. The built-in studio puts the script on a glass window high on the stage, near the lens, with two small ticks marking the line to read, so the gaze stays where a viewer expects it. Turn the prompter on, set the pace, and rehearse once with it running before a single take. If you prefer to speak without a metronome, switch on follow-my-voice and the text keeps up with you rather than the other way round; if you stop to think, it waits. Beam-splitter rig owners: flip the text, not the camera.
Record like takes are free, because they are
Countdown, slate, go. Do the whole clip even when you fluff a line; you will trim the front and back afterward and keep the best run. The review screen tells you if the audio was too quiet or clipping and if the shot was too dark or blown out, which are the two reasons most self-recorded video looks amateur, and both are fixable before the next take. If the clip needs your screen, share it mid-take and the camera drops into a corner the way it would on a call. Recording on a phone held upright gives you the vertical cut for free.
Save it to the library and put it to work
The take goes straight into your library, ready to be the clip a route step plays. Script the next branch and go again. Four branches and an opener is a full interactive pitch, and it is an afternoon’s work once the skeleton is in your head.
The script is not the enemy of sounding natural. The distance between the script and the lens is.
Two hundred and twenty words, one ask, eyes on the lens. Then press record.
One platform, seven layers
This article covered one slice. The machine ships whole: record it, get it found, arm it, test it, rank the leads, train the ads, and ask your AI how it is going.
Common questions
- How long should a sales video script be?
- For a ninety-second clip, about 220 words at a natural speaking pace: ten seconds of problem, twenty of cost, thirty for the turn, twenty of proof, ten for the ask. A 400-word draft is two videos, and the second is usually a branch.
- How do you read a script on camera without looking like you are reading?
- Keep the text near the lens so the eyes stay where the viewer expects, write short spoken sentences you have read aloud once, and rehearse with the prompter running before the first take. The built-in studio places the script high on the stage by the camera with eye-line marks and can follow your voice instead of scrolling on a timer.
- Do I need a teleprompter for a sales video?
- You need one if you want to say the same good thing every take. The studio prompter is built in and optional: turn it on, set the pace or let it follow your voice, and rehearse. Beam-splitter rig owners can flip the text only.
- Does StreamAgent write the script for me?
- No, by design. The studio helps you deliver a script you wrote, in your voice. Platforms spent this summer rewarding real people on camera, and a person reading a machine’s words is not the asset. AI in the product works around the video (transcripts, key moments, analysis), not in the frame.