How to Make a Product Demo Video People Actually Watch to the End

Most product demos are made the same way. Someone opens the app, hits record, clicks through every feature they're proud of, and stops eight minutes later. The video goes on the landing page, and almost nobody gets past the first minute.

The good news is that a demo people finish isn't about production budget. It's about a handful of decisions you make before, during and after you press record. This guide walks through them in order, with a script template and a checklist you can reuse.

Step 1: Give the video exactly one job

Before you write a word, finish this sentence: "After watching this, the viewer will ___."

• "...understand what the product does in 60 seconds" is a landing page demo.
• "...know how to set up their first project" is a tutorial.
• "...see that the new export feature exists and want to try it" is a feature announcement.
• "...be able to reproduce the bug" is a bug report.

Each of these is a different video, with a different length, structure and audience. The most common reason demos fail is that they try to do all four at once.

Rough length targets, as a starting point rather than a rule:

• Social feed clip: 15–60 seconds
• Landing page demo: 60–120 seconds
• Feature announcement: 1–2 minutes
• Tutorial: as long as the task takes, split into chapters

If your video is longer than its job needs, cut features, not speed. A fast-forwarded tour of everything is still a tour of everything.

Step 2: Write the script (yes, even for a 60-second video)

Unscripted demos ramble. A script doesn't have to be stiff, but it should exist. Spoken English runs at roughly 130–160 words a minute, so a 90-second demo is about 200–240 words. That's shorter than this section.

Here's a structure that works for almost any product demo:

1. The problem (5–10 seconds). Name the pain in the viewer's words. "Every time you ship a feature, someone has to record a demo, cut it, caption it and make a vertical version for social."
2. The promise (5 seconds). What changes. "Here's how to get all of that from one take."
3. The walkthrough (most of the video). Three steps at most. Show the action, then the result. Never narrate what's obvious on screen; say why instead of what.
4. The proof (5–10 seconds). The finished result, the number, the before and after.
5. The next step (5 seconds). One call to action. Not three.

Write it as you'd say it, then read it aloud once. Anything you stumble on, rewrite.

Tip: Put the script in a teleprompter that sits next to the app you're recording, so you're not glancing at another screen. Some recorders have one that follows your voice and pauses when you stop talking. ZumCast's teleprompter does this, and it never shows up in the recording.

Step 3: Set the stage

Your screen is the set. Ten minutes of preparation here saves an hour of editing later.

Clean up what's visible

• Close every app and tab you don't need. Hide the bookmarks bar.
• Hide desktop icons and turn on Do Not Disturb so notifications don't pop in.
• Use demo data. Real customer names, emails and invoices have no place in a public video (more on this below).

Make things readable

• Record a single window at a fixed size, such as 1920×1080 or 1280×720, rather than a 5K display. Viewers watch in a small player, often on a phone.
• In a browser, zoom in to 125–150%. Text that looks huge to you looks normal in the video.
• Make the cursor bigger in your Mac's accessibility settings, or use a recorder that enlarges and smooths it afterwards.

Get the sound right

• Use an external mic or a headset if you can. Viewers forgive average video far more easily than bad audio.
• Record in a small room with soft furnishings. Bare walls echo.
• Do a ten-second test recording and listen back with headphones before the real take.

Step 4: Record in sections, not one heroic take

You don't need a perfect take. You need clean pieces.

• Record one section at a time, following your script. It's much easier to redo 20 seconds than four minutes.
• If you fluff a line, pause, then start the sentence again. Don't stop recording. A clean pause is easy to cut.
• Move the mouse deliberately. Go straight to the target, pause on it, click. Wandering cursors make people seasick, especially once zooms are added.
• Wait a beat after each click so the result is on screen before you talk over it.

If you're on camera, put the camera at eye level and look at it during the problem and the call to action. During the walkthrough, look at the screen like the viewer does.

Step 5: Edit for attention

This is where most of the watch time is won or lost. In order of impact:

1. Cut every pause and filler word. Viewers notice dead air far more than you do. A three-second pause on screen feels like ten.

2. Zoom in on every important click. A full-screen view of a dense app tells the viewer nothing about where to look. A smooth zoom onto the button you're about to press does. Ease in, hold, ease out; don't jump.

3. Make the cursor obvious. A larger cursor, smoothing to steady shaky movement, and a subtle click effect.

4. Add captions. Many people watch with the sound off, especially in social feeds, and captions also help non-native speakers and anyone in a noisy place. Burn them into feed videos; upload a caption file (SRT or VTT) for YouTube so viewers can turn them off.

5. Add a title card and an end card. Keep the title to a few words. End on the one call to action from your script.

6. Watch it once at full speed, with the sound off. If it doesn't make sense muted, the visuals aren't doing enough.

Doing all this by hand in a general-purpose video editor takes a while: every zoom is a keyframe, every cut is a scrub. Purpose-built screen recorders now do the mechanical parts automatically. In ZumCast, zooms on every click and a smooth cursor are already on the timeline when you stop recording. You can delete words from the transcript to cut the video, or ask Zumi, the built-in AI editor, in plain words: "cut the pauses", "zoom in on the Save button", "add captions at the top". Zumi shows the plan on the timeline first, and nothing changes until you press Apply. It runs on your Mac, so your unreleased product never leaves it.

Step 6: Make one take into every format

The same demo needs to live in several places, and each wants a different shape:

• Website, YouTube, docs: 16:9 at 1080p, the master version.
• YouTube Shorts, Instagram Reels, TikTok: 9:16 vertical. Don't just crop the middle; reframe so the action stays in view.
• LinkedIn feed: 1:1 or 4:5 takes up more of the feed than 16:9.
• Docs, GitHub READMEs, help centres: a short, silent GIF of one action.
• Help articles and onboarding: a step-by-step written guide with a screenshot per step.

Re-recording for each format is how a one-hour job becomes a one-day job. Instead, record once and derive the rest. A vertical version needs its own crop that follows the action, not a fixed centre crop. A written guide can be generated from the clicks in your recording, one step per click.

ZumCast keeps each format as a version of the same project: 16:9, 9:16, 1:1 and 4:5 each with their own crop. Export presets cover YouTube, X, LinkedIn, Shorts and Reels, and Export all writes every version in one go. The same take also becomes a step-by-step guide (Markdown, HTML or PDF) and a clickable interactive demo.

Step 7: Title, thumbnail and chapters

• Title: say what the viewer gets, not what the feature is called. "Turn one screen recording into a vertical clip" beats "Introducing Versions 2.0".
• Thumbnail: one big, readable word or number on a frame from the video. Test it at the size of a postage stamp.
• Chapters: for anything over two minutes on YouTube, add timestamps in the description so people can jump to the part they need.

Step 8: Publish, then look at where people leave

YouTube Studio's audience retention graph shows the exact second people drop off. Watch your video at that moment. It's almost always one of three things: a slow intro, a long pause, or a section that isn't what the title promised. Fix it in the next video. The second demo you make with this feedback will be noticeably better than the first.

The checklist

Before recording

• One job for this video, in one sentence
• Script of about 150 words per minute of video
• Demo data, no real customer details
• Clean desktop, notifications off, one window at a fixed size
• Browser zoomed to 125–150%
• Mic tested with headphones

While recording

• One section at a time
• Pause and restart a sentence instead of stopping
• Slow, direct mouse movement; wait a beat after each click

When editing

• Pauses and fillers cut
• Zoom on every important click
• Cursor visible, captions on
• Title card and one call to action
• Watched once with the sound off

When publishing

• 16:9, 9:16 and 1:1 versions where you need them
• Caption file uploaded on YouTube
• Clear title, readable thumbnail, chapters for long videos

---

ZumCast is a Mac screen recorder with an AI editor that runs on your Mac. It adds zooms on every click, cuts pauses and fillers, writes captions and makes the vertical clip, GIF and step-by-step guide from the same take. The Basic version is free; Pro is a one-time purchase. Download it free.