Using a Teleprompter for Product Demos: How to Script and Record a Demo Video That Matches What's on Screen

A product demo is the hardest kind of video to read from a script, and the kind that most needs one. A talking-head intro has to be tight and confident, because it decides whether anyone watches the next ninety seconds. The middle is screen recordings or shots of the product, where the words have to land at the exact moment the cursor moves or the lid opens. And the ending has to say, precisely, what to do next. Wing any of those and the demo gets re-cut three times; script all of them and the talking-head parts sound like a hostage video.

This post is a walkthrough for recording a product demo on an iPhone with Teleprompter: Camera Overlay: a structure that works for a SaaS walkthrough, an app demo and a physical product, how to write the script so it matches the screen, which scrolling mode to use for which segment, and the parts of a demo where you should put the teleprompter down. It builds on the sales and outreach videos post, which covers the one-to-one version, and the online courses and tutorials post, which covers longer instructional recordings.

The two kinds of footage in every demo

Before writing anything, split the demo into its two kinds of footage, because they are recorded differently and scripted differently.

• Face segments. You, on camera, talking to the viewer: the hook, the problem, the transitions and the call to action. These are recorded with the teleprompter overlay next to the front lens, so you read while keeping eye contact.
• Show segments. The screen recording, the app on a second phone, the product on a table. Your voice is a voiceover here, recorded either live while you demonstrate or separately afterwards, and the teleprompter is a voiceover prompter rather than an eye-contact tool.

Most bad demos fail because they treat the whole thing as one segment. The face segments get recorded in one long take that includes a lot of 'and then you click here', which the viewer cannot see, and the show segments get narrated from memory, so they ramble. Mark each line of the script with F or S before you record and the problem mostly disappears.

A five-part structure that fits two minutes

The structure below is the one that survives contact with a real product. Each part has a job, a kind of footage and a rough word budget at a conversational 140 words a minute.

• Hook (F, 10 to 15 seconds, about 30 words). One sentence about the outcome, not the product. 'This is how I close the books in twelve minutes instead of a day' beats 'Today I'm going to show you our accounting dashboard.' The hooks post has a dozen patterns.
• Problem (F, 15 seconds, about 35 words). The annoying thing the viewer does today. Name it specifically enough that they nod.
• Three features (S, 60 to 75 seconds, about 150 words). Three, not seven. Each one is: what you do, what happens, why it matters. The words must match the screen, which the next section covers.
• Proof (F or S, 10 to 15 seconds). A number, a before-and-after, or a short line about who already uses it. Only claims you can back up.
• Call to action (F, 10 seconds, about 25 words). One action. Free trial, link below, book a call. Not all three.

Written out, that is 250 to 300 words, which is a two-minute demo. If your draft is 500 words, the editing down by 30% post has the cutting method; demos nearly always lose their fourth and fifth feature first.

Writing lines that match what is on screen

The show segments are where product demos are won. The rule: every sentence describes something the viewer can see at that moment, in the order they see it. Write the script in two columns, action on the left, words on the right, and do not let a sentence outrun its action.

• Lead with the verb the viewer would use. 'Drag the invoice onto the month' rather than 'The system supports drag-and-drop invoice allocation.'
• Say what happened immediately after it happens. 'It picked up the supplier and the amount' as the fields fill in, not three sentences later.
• One idea per screen state. If you have to say two things about one screen, show it twice, or cut one.
• Avoid 'as you can see.' If they can see it, describe what it means; if they cannot, you need a closer shot, not a narration.
• Write the UI words exactly as they appear. If the button says Reconcile, say reconcile, not 'match them up.'

Keep the face segments in the voice you used in the write a script that doesn't sound written post: short sentences, contractions, one breath per line. The show segments can be slightly more clipped, because the picture is doing half the work.

Which scrolling mode for which segment

Teleprompter: Camera Overlay has two ways to move the script, and a demo uses both. The voice-follow vs auto-scroll post explains the mechanics; this is how they map onto a demo.

• Face segments: voice-follow. The script floats over the camera preview next to the front lens and scrolls as you speak, using the iPhone's on-device speech recognition. You can pause to think, restart a sentence, or jump back to the top, and it finds your place again. Eye contact stays on the lens because the words are next to it.
• Show segments recorded live: voice-follow. When you are clicking through the product on a laptop while the iPhone records your voiceover, you will pause constantly to wait for a page to load or a modal to open. Voice-follow waits with you. Auto-scroll does not, and you end up chasing the text.
• Show segments recorded as a fixed-length voiceover: auto-scroll. If the screen recording is already cut to exactly 48 seconds and you are laying a voice track over it, set the speed slider so the script takes 48 seconds, tap to pause if needed, and read it like a radio spot. The constant pace is what you want here.
• Either mode: adjust the text size up for show segments, where you are not looking at the lens and can afford bigger lines, and down for face segments, where a narrow column keeps your eyes close to the camera.

In both modes the script never appears in the saved video; you get a clean 4K portrait clip to drop into the edit next to the screen recording. Everything runs offline, so a demo recorded in a meeting room with no Wi-Fi works the same.

Recording order: all the faces, then all the screens

Do not record the demo in viewer order. Record every face segment back to back in one sitting with the same light and framing, then every show segment. The batch-recording post explains why this halves the time; for a demo there is an extra reason: the face segments have to cut together as if they were one take, and they will only match if they were.

• Save each segment as its own script in the app (hook, problem, proof, CTA), so each take starts at the top of the right text instead of scrolling to the middle of a long one.
• Record two or three takes of each and use the takes library to pick later; do not judge takes while recording.
• Say the first word of the next segment at the end of each face take, then cut it off in the edit. It makes the transitions sound continuous.
• For show segments, record the screen first, then narrate to it. Narrating while clicking is only worth it when the pauses are part of the point.

When not to use a teleprompter in a demo

There are three parts of a demo where reading hurts more than it helps, and it is worth being honest about them.

• Live-click walkthroughs where the product is unpredictable. If the thing you are demonstrating does not always respond the same way, a script will be wrong half the time. Use a bullet list of the three points per screen and talk, then tighten in the edit.
• Q&A and objection handling. A viewer can hear the difference between an answer and a recited answer. Script the question, improvise the reply.
• Very short feature teasers under 20 seconds. By the time you have set up the overlay, you could have said it. Memorise the two lines.

The teleprompter earns its place in the hook, the problem, the proof and the CTA, which are also the four parts that decide whether the demo works. Script those tightly, read them with the overlay next to the lens, and let the product speak for itself in the middle.

Where the series goes next

Product demos were the last topic on the list this series started with: from writing a script that doesn't sound written, through platform guides for Reels, Shorts and TikTok, to setup, mistakes, readability, compliance wording, speeches and editing. With the bank complete, the next posts come from what readers actually search for. The first of those is a troubleshooting post, because 'why is the teleprompter not visible' is a question people keep asking: the overlay not showing, text hidden behind the camera controls, mirror mode flipping it, and the other display problems with their fixes.