Open the landing page of almost any AI or fintech product launched this year and the top of the page moves. A question types itself into a chat box. Cards fan out. A button gets pressed by a cursor that isn't yours, results slide in, a number counts up, the name lands. Thirty to sixty seconds, no voice or a quiet one, music underneath, and it loops.
That is a hero video. It isn't a demo and it isn't a tutorial: nobody learns how to use the product from it. Its job is to make a visitor understand in one glance what the product does and want to scroll down.
Most teams think they need a motion designer for this, or at least a polished screen recording. You need neither. This guide covers how to plan one, what to put in each scene, how to write the words, the small motion details that separate "expensive" from "template", and three scene-by-scene templates you can copy. The last section shows how to make one in Nook from your website in a few minutes.
What a hero video is (and isn't)
A hero video is the moving picture at the top of a landing page, a launch post or an app store listing. It has three jobs, in order:
- Say what the product does in the first five seconds, to someone who hasn't read a word.
- Show the moment of value: the question answered, the task done, the problem caught.
- Leave one thing to do next: start free, book a demo, join the waitlist.
It is not:
- A demo video. A demo walks through the real product, click by click, usually narrated. It belongs further down the page or in a sales email. See how to make a product demo video.
- A walkthrough. A walkthrough teaches one task. It belongs in the help centre. See walkthrough videos for business software.
- A brand film. No actors, no drone shots, no mission statement read over slow piano.
The difference matters because it changes what goes on screen. A demo shows your real interface. A hero video shows a simplified, drawn version of it: the one input, the one button, the one result that matter, at a size anyone can read on a phone.
Why you don't need a screen recording
Real product screens make poor hero video material. They're dense, the text is small, the colours are your app's and not your brand's, and the interesting part is buried in a corner. That's why the hero videos you admire draw their interfaces instead of recording them: a clean chat box, three cards, a button, a short result. Every element is legible at a glance and moves exactly when it should.
Drawing also means the video doesn't go stale when your UI changes, and you can make one before the product is finished. A drawn chat box from a pre-launch landing page is still accurate after the redesign.
If you do have good screenshots, use one or two, framed in a browser or a phone, for the moment where the real product says more than a drawing. One or two, not ten.
The shape that works: five beats in 45 seconds
Almost every good hero video follows the same arc. Give each beat one to three scenes, and each scene one idea.
| Beat | Time | What it does | Typical scenes |
|---|---|---|---|
| Hook | 0–5 s | Names the problem or the promise in a few words | A big headline, stacked lines |
| Ask | 5–12 s | Shows what the user brings: a question, a document, a task | A question typed into a chat box, a command palette |
| Answer | 12–30 s | The product does the thing | Results sliding in, an answer streaming, an agent ticking off tasks |
| Proof | 30–40 s | Why to believe it | A number counting up, a customer quote, before and after |
| Close | 40–45 s | One action and the name | A call to action with a button, the logo |
A 30 second video drops most of the proof. A 60 second one adds a second ask-and-answer pair, usually a second use case or a second kind of user.
Two rules keep it watchable:
- One idea per scene. If a scene needs a sentence to explain, it's two scenes.
- About two seconds per idea. A headline gets two seconds. A typed question gets three or four. A scene with results gets four. Nothing gets more than six unless the viewer has to read a quote.
The scenes, and what each one is for
Think of these as a library. You rarely need more than eight of them in one video.
Words
- Headline. Two short sentences in big type, one per line: "Spreadsheets. Everywhere." It's the spine of the video; use two or three, for the problem, the promise and the result. Colour the one word that matters: "Answers. Not reports."
- Stacked lines. Two to four lines landing one after another, the earlier ones dimming as the next arrives. "Receipts in one place. / Spend in another. / Answers in neither." A strong opening because it builds tension.
- Changing word. "Built for finance / ops / founders teams", the middle word sliding out as the next slides in. Good for showing who it's for without three separate scenes.
- Big number. One number counting up with a label: "3× faster month-end close". Only use a number you can stand behind.
- Quote. A customer's words with their name and role. Only a real one.
Your product at work
These are the scenes that make a hero video feel like your product rather than a stock template.
- Typed question. A chat or search box where a question is typed and sent. Write what a real user would type, in their words: "Why were we 40% over budget on travel?", not "Analyse my expenses".
- Question and answer. The question sent, a moment of thinking, then the answer streaming in a few words at a time, with its sources underneath. The defining scene for an AI product.
- Agent at work. A task list working itself through: each line spins, then ticks. "Match 1,284 receipts. Flag 3 invoices. Draft the report. Send it." Shows that the product does the work, not just answers.
- Command palette. ⌘K, a word typed, the list narrowing, the right line picked. Says "fast" and "keyboard-first" without a word.
- Button and results. A cursor presses "Check for anomalies"; two or three result cards slide in, one flagged, one passed.
- Suggested edit. A paragraph with a change suggested in red and green, then accepted. For writing, legal and review tools.
- Form. Fields filling themselves in, then sent. For anything that replaces data entry.
- Board. A card dragged from "To review" to "Done". For workflow tools.
- Working together. A document with two or three named cursors moving over it and a comment popping up. For anything collaborative.
Lists and data
- Cards. Two to four small cards fanning in, each a thing the product works with: invoices, bank feed, calendar.
- Checklist, steps, before and after, chart, notifications, settings switches, a table. Use these for proof and "how it works", one at a time.
Your media
- A screenshot in a browser, phone or laptop frame, pushing in slowly.
- A marked-up screenshot: a circle, arrow or spotlight drawn on the one spot that matters, then the view pushes in on it.
- A short clip of the real product, when nothing else will do.
Writing the words
Hero video words are closer to billboard copy than to a script.
Headlines: two to six words, full stops. "Anomalies. Detected." reads faster than "Detect anomalies automatically". The full stop is what makes it land as two beats.
Typed questions: the customer's words. Read your support inbox or sales call notes and copy a real question. Specific beats general: "Which invoices are still unpaid this month?" says more about the product than "Ask your data anything".
Results: concrete and in your domain. "3 line items on the invoice aren't on the purchase order" is a result. "Insights generated" is not.
Numbers: only true ones. If you don't have a number yet, use a before and after or a checklist instead. Made-up example values inside the drawn interface (a sample invoice amount, a vendor name) are fine, because nobody reads them as claims.
The close: one verb. "Start free", "Book a demo", "Join the waitlist". Not two buttons.
If there's a voiceover, budget about two words per second of scene, with a breath either side. A 3 second scene takes six words at most, or stays silent. Headline scenes usually stay silent: the words are already on screen. Never open with "Introducing…" or the product name.
The motion details that make it look expensive
The difference between a hero video that looks designed and one that looks like a template is almost never the scenes. It's how things move. Six details do most of the work:
- Text is typed like a person types. An uneven rhythm, a pause at a comma, a slightly longer one at a space now and then. A perfectly steady typewriter looks like a robot. Even better, show a correction: a slip rubbed out, or a word selected and typed over ("40%" becomes "$8,400"). It's the single most human detail you can add.
- The caret behaves. Solid while typing, blinking only when typing stops. Real editors do this, and viewers notice the wrong one without knowing why.
- The cursor changes shape. An arrow over empty space, a pointing hand over the button it's about to press, a closed hand while dragging. And it never sits perfectly still; a few pixels of drift while it waits.
- Lists tighten as they go. The second item arrives a little after the first, the third a little sooner after the second, rather than all at an even beat. Even spacing reads as mechanical.
- Fades are short, movement is long. Things should appear quickly and settle slowly, with the settling about three times longer than the fade. Children land a beat after their parent: the tick arrives after its row.
- Nothing is ever quite still. A slow push in under the whole scene, a percent or two, keeps a held frame alive.
And transitions: scenes on the same background push each other aside; where the background changes, let the old scene settle away and the new one land, with a frame of nothing in between. Use one transition style for the whole video.
Look and sound
Colour. Your brand colour as the main backdrop, near-black for the big headlines, off-white for the scenes with cards. Don't put two dark scenes in a row.
Type. One headline font, one interface font. A serif headline over sans-serif UI is the current look for AI and finance products; all-sans reads more technical.
Music. Calm and steady under 45 seconds of motion. The music should duck under any voice line.
Captions. Most hero videos play muted, on autoplay. If there's a voiceover, burn in captions. If there isn't, the words on screen must carry everything.
Logo. Small in a corner is enough. The name gets its own scene at the end.
Sizes, length and where to put it
| Where | Size | Length | Notes |
|---|---|---|---|
| Landing page hero | 16:9 | 30–45 s | Autoplay, muted, loop. Poster frame of the first headline. |
| Product Hunt, launch tweet | 16:9 or 1:1 | 30–45 s | The first two seconds must work as a still. |
| LinkedIn, Instagram, TikTok | 9:16 or 1:1 | 30 s | Tall needs bigger type; check every scene on a phone. |
| Sales email | GIF of the first 5–8 s, linked | — | Video doesn't play inside most email clients. |
| App store listing | The store's size | 15–30 s | Check the store's rules on drawn versus real screens. |
On the landing page itself:
- Autoplay muted and loop, with no controls in the way. A visitor who wants sound can click.
- Set a poster image (the first headline works well) so the hero isn't blank while the video loads.
- Keep the file small. A 45 second hero at 1080p is a few megabytes as an MP4. Lazy-load anything below it.
- Put the call to action next to the video, not only inside it. The video makes them want it; the button beside it is what they click.
The ten mistakes we see most
- Opening on the logo. Nobody knows you yet. Open on the problem.
- Too many scenes. Twelve scenes in 30 seconds is a slideshow. Eight is plenty for 45 seconds.
- Your real UI, full size. Unreadable at hero size. Draw the one part that matters.
- Generic questions. "Ask anything" says nothing. Use a real customer's words.
- Every scene the same speed. Vary it: a quick headline, a slower answer, a held number.
- Robotic typing. Steady, perfect typing reads as fake. Add rhythm and one correction.
- Numbers you can't back up. One made-up "10× faster" undoes the rest.
- Two calls to action. Pick one.
- Voice reading the screen word for word. If the words are on screen, the voice should add to them or stay quiet.
- Never checking it on a phone. Half your visitors will see it at a third of the width you designed it at.
Three scene-by-scene templates
An AI assistant (45 s)
- Stacked lines (3.5 s): "Receipts in one place." / "Spend in another." / "Answers in neither."
- Typed question (4 s): "Why were we 40% over budget on travel?", with "40%" selected and typed over as "$8,400".
- Question and answer (5 s): the answer streams in, two sources underneath.
- Agent at work (4 s): four tasks tick off; "Done in 42 seconds".
- Command palette (3.5 s): "unpaid" typed, the list narrows, Enter.
- Changing word (3 s): "Built for finance / operations / founders teams".
- Call to action (3 s): "Ask your books anything", button "Start free".
- Name (2 s).
A workflow or collaboration tool (40 s)
- Headline (2 s): "Four tools. One job."
- Cards (3 s): the four tools it replaces.
- Working together (3.5 s): three named cursors in one document, a comment pops up.
- Board (3.5 s): a card dragged to "Done", its tag turns green.
- Notifications (3 s): the update arrives where people already are.
- Before and after (4 s): "9 days to close" crossed out, "2 days" ticked.
- Call to action (3 s): "One place. Done." button "Start free".
- Name (2 s).
A developer tool (40 s)
- Headline (2 s): "Deploys. Without the yaml."
- Code (4.5 s): four lines typed into a terminal, the last one green.
- Steps (3.5 s): init, deploy, watch.
- Settings (3 s): "Preview every branch" and "Roll back in one click" switching on.
- Big number (3 s): "40s from push to live", only if it's your median.
- Call to action (3 s): the install command as the button.
- Name (2 s).
How to make one in Nook
Nook's video builder makes exactly this kind of video, from a library of the scenes above, without a recording.
- Start from your website or a few lines. Paste your address (or describe the product, or add a few screenshots) and Nook writes a first draft: the scenes, the words in them, your colour and the backdrop for each. Or start from a template: launch, feature announcement, comparison, social proof, developer tool, AI product, or blank.
- Change anything. Every scene has its own words, backdrop and length; add scenes from the library, move them, delete them. Typed questions can be typed like a person, a word at a time or pasted, with a correction, a rewrite, an autocomplete or an @mention after. The preview plays live as you edit, in wide, tall or square.
- Add a voice if you want one. Have the script written for you, line by line, each measured against its scene, or record it in your own voice. Put yourself, or a presenter, in a bubble in the corner.
- Make it. Nook draws the final video at full quality, with captions, music and your logo, ready to put at the top of your page.
Writing drafts, editing and previewing are free. Making the final video takes a plan, from $9 a month. You can see one Nook made of itself, from usenook.live, on our home page, and start your own in the video builder.
If you do have a screen recording and want a launch video built over your own clicks instead, see how to make a product launch video.
Questions
How long should a SaaS hero video be?
30 to 45 seconds for a landing page. Long enough for the five beats, short enough to loop without anyone noticing it started again. Sixty seconds works when you have two distinct use cases to show.
Does a hero video need a voiceover?
No. Most play muted, so the words on screen must carry the story either way. A voiceover helps when the video is also shared on its own (a launch post, a sales email), and then it needs captions.
Should I use my real product screens?
Sparingly. Draw the interface for most scenes, so it's legible and on brand, and use one or two real screenshots in a frame where the real product says something a drawing can't.
Can I make one before the product is finished?
Yes, and it's one of the best uses. A drawn chat box and result cards show the promise without depending on a finished UI, which is why pre-launch waitlist pages use them.
What's the difference between a hero video and a product demo video?
A hero video shows what the product does in under a minute, drawn and simplified, to someone who has never heard of it. A demo shows how it works, on the real product, to someone already interested. Most landing pages want both: the hero at the top, the demo further down.
How often should I update it?
When your positioning changes, not when your UI does. Because the interface is drawn, a hero video stays accurate through redesigns; a new headline or a new use case is the usual reason to make a new one.
Make your first demo video free
Upload a screen recording. Nook writes and voices the script, adds captions and writes the step-by-step guide.
Make a free video