To film a 90-second product demo that converts cold visitors, show one kind of person finishing one job in your real product: flash the finished result in the first five seconds, say who it is for, do the job once on screen, land on the success screen, and close with the same call to action as the button next to the player. Record it with Loom or any screen recorder and a mic close to your mouth, make sure it reads with the sound off, and judge it after 14 days by whether visitors who pressed play signed up more often than visitors who didn't.
Last updated: Nov 10, 2026
This is a solo-maker playbook for the demo that sits next to your signup button, not a brand film. It covers a one-job brief, a second-by-second script, a cheap recording setup, captions that survive muted autoplay, placement, launch reuse and a two-week check. Research figures and platform rules link to their sources (read 10 November 2026). The worked example near the end uses a made-up product and is illustrative.
Why 90 seconds and not four minutes?
A stranger on your landing page has one question: is this for me? Most founder demos answer it late. They open with an origin story, move on to architecture, then tour every feature that took a weekend. Cold traffic never asked for any of that. It asked whether the tool finishes the job already in its head.
Ninety seconds forces the cut. Roughly:
0 to 15 seconds: the result, who it's for and the painful job
15 to 70 seconds: the product doing that job once
70 to 90 seconds: what "done" looks like and one clear next step
Longer videos earn their place in docs, sales follow-ups and YouTube tutorials, where the viewer is already invested. On a page for strangers, every extra minute is a tax. If you can't show one complete win in under two minutes, the product story is still mushy, or you're demoing the wrong job.
Before you script, get one outsider to react to a rough cut. The same people who will give you brutal feedback before you launch publicly will tell you within ten seconds whether they know what the product is for.
What does the data say about demo length and play rate?
Wistia publishes benchmarks from the videos hosted on its platform. Its 2026 State of Video report (PDF) draws on more than 13 million videos, and three of its tables matter for a landing-page demo.
Wistia 2026 benchmark | Figure | What it means for your demo |
|---|---|---|
Engagement (average share watched) for product demo videos under 1 minute | 50% | Even short demos lose half their audience along the way. |
Engagement for product demo videos of 1 to 3 minutes | 46% | At 90 seconds, the average viewer sees roughly the first 40 seconds. Put the result before that point. |
Engagement for product demo videos of 3 to 5 minutes / 5 to 30 minutes | 40% / 28% | The drop gets steeper as demos get longer. |
Play rate on homepages | 24% | About a quarter of homepage visitors press play. |
Play rate on product pages | 15% | Fewer visitors press play once they're deeper in the site. |
Play rate on pages Wistia classes as landing pages | 6% | Most visitors never watch, so the page has to convert without the video. |
Two habits follow. First, show the ending first. Wistia's own advice in its guide to marketing video length is to put the most important information in the first half, because most viewers drop off around the halfway point. Second, protect the opening seconds. An older Wistia analysis of audience retention (2020) found the average drop in the first 2% of a video was 4.9% for videos of 1 to 2 minutes and 17.3% for videos of 5 to 10 minutes. Its fixes for that early drop are blunt: shorter intros, clear audio, no logo sequences, and proving the point instead of stating it.
Treat these as rough benchmarks across many industries, not targets for your product. Your own 14-day numbers (below) are the ones that decide anything.
What should you decide before you open the camera?
Don't start recording. Fill a half-page brief first. If you can't fill it, you aren't ready for a demo. You're ready for another positioning pass.
One ICP sentence: "Solo founders running a B2B SaaS with under 50 customers." Not "everyone who wants to grow."
One job: a verb and an object they already understand. "Turn a messy Notion dump into a launch checklist." Not "AI-powered productivity."
One success screen: the exact UI state that means it worked. Export ready, invite sent, first report green. Pick one.
One CTA: signup, book a call or start a trial. Match the button on the page. Don't invent a third path in the video.
What you won't show: integrations, settings, pricing tiers, the roadmap. Write them down so you don't "just quickly" open them.
Keep the brief next to your mic. When you drift into a feature tour mid-take, stop and re-record from the job.
What does a 90-second demo script look like, second by second?
Here is a template you can fill in. The columns force every spoken line to have something on screen and a caption that still works muted. Times are targets, not a metronome.
Time | On screen | What you say | Caption or on-screen text |
|---|---|---|---|
0:00 to 0:05 | The finished result: the report, the sent invite, the green status | "This is [result], done in [time] with [product]." | The outcome in under eight words |
0:05 to 0:15 | Your face or a still of the messy "before" | "If you're a [ICP] who [painful job], this is for you. It's not for [who it isn't for]." | "For [ICP]" |
0:15 to 0:25 | The before state inside the product: blank project, raw CSV, cluttered inbox | "Here's where most people start." | "Start: [before state]" |
0:25 to 0:45 | Steps one and two, zoomed so text is readable on a phone | Narrate the decision, not every click: "I paste the brief and pick the one goal." | Three to six words per step |
0:45 to 1:05 | Step three and the result appearing | "Now it builds the [thing]." Then stop talking for half a second. | The step name |
1:05 to 1:20 | The success screen with the proof highlighted | One sentence on what "done" means for them. | "Done: [result]" |
1:20 to 1:30 | The page's real button or URL | "If that's your job, [CTA] from the button on this page. You'll reach this same screen in [honest time]." | The CTA text, word for word |
Speak like someone who uses the product, not like a trailer voice. Three lines worth stealing:
Hook: "If you're trying to [job] without hiring [expensive alternative], here's 90 seconds of how [product] does it."
Who it's not for: "This isn't for enterprise teams with a full ops stack. It's for [ICP] who need [outcome] this week."
CTA: say it once, then stop. No outro music, no "like and subscribe."
Read the script aloud once at conversation speed with a timer. If it runs past 100 seconds, cut adjectives and the weakest step, not the success screen. Don't fix it by talking faster.
How do you record it with Loom or a phone?
You don't need a studio. You need a clear voice track, a readable UI and audio that doesn't sound like a bathroom.
Loom or a similar screen recorder: the best default for SaaS. Screen with a face bubble, or screen only with voice. Record at 1080p in a quiet room, and zoom the browser so text is readable on a phone.
Phone on a stack of books: fine for mobile-first products or a talking-head intro cut with screen footage. Shoot landscape unless the product is mobile only. Lock exposure and wipe the lens.
Mic: any lavalier or USB mic beats a built-in mic three feet away. With only earbuds, sit closer and switch off the fan or AC.
Screen hygiene: a clean desktop or a dedicated browser profile, believable demo data, no Slack badges, no calendar popups, and a larger system font if your UI is dense.
Do three takes at most per session. Take two is usually the one. If take three still rambles, the brief is wrong: rewrite the job instead of hoping to fix it in the edit.
One thing to know before you pick a tool: captions differ a lot by host. Loom's help page on transcripts and closed captions says transcripts come on all plans and viewers can toggle captions in the player, but downloading the captions as an SRT file is limited to Business, Business + AI and Enterprise plans. A Loom embed is fine for a quick share. If the video will autoplay muted on your homepage, you'll want captions burned into the file or a player that shows a caption track by default.
How much should you edit?
Very little. Cut dead air at the start and end, remove the one cough, and keep speed at 1.0x unless you speak painfully slowly (then 1.1x at most). Skip whoosh transitions, stock music beds and logo stingers that eat eight of your ninety seconds.
Export an MP4 at 1080p, and keep it under about 25 MB if your CMS is picky. Name the file after the job, not "final-final-v7."
How do you make the demo work with the sound off?
Plenty of visitors will meet your demo muted. Browsers decide that, not you. Chrome's autoplay policy says muted autoplay is always allowed, while autoplay with sound needs the visitor to have interacted with the site or to meet other conditions. WebKit's video policies for iOS let a video autoplay without a tap only when it has no audio track or is muted, and on iPhone it plays inline only with the playsinline attribute. So if you autoplay, you're autoplaying a silent film.
Captions are also an accessibility requirement, not decoration. WCAG success criterion 1.2.2, Captions (Prerecorded) is Level A, the baseline level, and its definition of captions covers meaningful non-speech sound as well as dialogue. It also notes captions shouldn't cover relevant information in the video, which for a demo means the UI you're pointing at.
Use this checklist before the video goes on the page:
Every key line in the script appears as a caption or on-screen text.
Captions sit where they don't cover the button or field you're demonstrating.
Auto captions were reviewed by hand. YouTube's own help on automatic captioning says they can misrepresent speech and should always be reviewed, and that a long silence at the start of a video is one reason they may not generate at all. Start talking in the first seconds.
Product names and jargon are spelled correctly in the captions.
If the player autoplays, it has
mutedandplaysinline, a visible play or unmute control, and no pause icon showing when nothing is playing (Chrome's guidance is never to assume a video will play).The first frame works as a still, in case the video never starts.
You watched the whole thing once on a phone, muted, and still understood the job.
Where should the video go, and what should the thumbnail show?
A demo nobody presses play on is a blog post you filmed by accident. Wistia's 6% play rate for landing pages is the reminder: the headline, screenshot and button have to carry the page on their own, and the video is for the visitor who is interested but not yet sure.
Thumbnail: one readable UI win, plus your face if you have trust to borrow, or a clear "90-sec demo" label. Show the success screen, not your logo wall. The same rules as product screenshots that get clicks apply: one outcome moment, real-looking data, text big enough for a small tile.
Above the fold: put the player next to the primary CTA, not under six feature sections.
Poster frame: never a blank countdown. Start on the result, a title card or the before state.
YouTube hosting: per YouTube's help on custom thumbnails, you need a verified account to upload one, and it should be 16:9 with a minimum width of 640 pixels. It is also the preview image in the embedded player.
Mobile: check the player on a phone. If UI text is unreadable, re-record with zoom or a simpler path.
Pair the video with honest proof nearby: one quote, one number you can defend, not a fake logo strip. If you're still thin on proof, use the same discipline as writing social proof for an indie launch with almost no users: permissioned quotes beat invented authority.
Remember the demo sells the start of the job, and activation still has to finish it. A great video that leads into an empty dashboard is a broken promise, so fix the first ten minutes the way you'd stop obsessing over the landing page and fix time-to-first-action.
How do you reuse the demo for a launch listing?
The same cut usually works on launch day. Product Hunt's guide to building your product page calls video optional but recommended, and says it doesn't need to be polished: a simple screen recording with your voice can do more than a big-budget trailer. Video or demo sits in the "Make it stand out" part of its pre-launch checklist, next to three or more gallery images at a recommended 1270x760px.
Before you submit anywhere, run the page the demo sits on through the free MakerHunt launch readiness checker. It flags missing titles, meta descriptions, Open Graph images and indexing problems, which a stranger meets before they ever see your video.
If your product changes every week, plan for re-shoots. A demo showing last month's navigation erodes trust fast. Script around the job rather than the menu, and keep the surrounding copy modular, as in writing a landing page that still converts when your product changes weekly.
What does this look like for a real product?
Illustrative example. The product, timings and numbers below are made up to show the method.
Take "Shelfnote," an imaginary tool that turns supplier invoices into a reorder list for small shop owners. The founder's first demo ran 3 minutes 40 seconds: a founding story, a settings tour, the supplier integrations page and, at 2:50, the reorder list.
The rewrite started with the brief. ICP: owners of a single shop with fewer than 200 products. Job: know what to reorder this week without a spreadsheet. Success screen: the reorder list with quantities filled in. CTA: "Start free," matching the hero button. Won't show: integrations, pricing, multi-store.
The new script:
0:00: the finished reorder list on screen. "This is next week's reorder list, built from three supplier invoices in under a minute."
0:06: "If you run one shop and still reorder from memory, this is for you."
0:16: a folder of three PDF invoices dragged into an empty Shelfnote workspace.
0:28 to 1:02: line items appear, one low-stock item is flagged, and the founder adjusts one quantity.
1:05: the finished list, cursor circling the total. "That's the order. Send it or print it."
1:22: "Start free from the button on this page and you'll see your own list in about ten minutes."
It ran 1:31 on the second take. Captions were burned in, positioned at the top of the frame because the action happened at the bottom of the screen. Over the next 14 days, the founder's sheet showed (illustrative) 2,100 landing sessions, 190 plays (about 9%), and a signup rate of 11% among sessions that played against 3% among those that didn't. That split doesn't prove the video caused signups, since curious people both press play and sign up more, but it was enough to keep the video and move on to the next leak: two of five replies said the free plan's limits weren't clear.
How do you know whether the demo is working?
Ship the demo, then track a short list for two weeks. If you can't instrument it, a spreadsheet and daily manual counts still beat vibes.
Play rate: plays divided by landing sessions. Are people pressing play at all?
Watch depth: the share who reach about 75%. Did the job hold their attention?
Watch to signup: the signup rate of sessions that played against those that didn't. Even a rough split is gold.
Qualitative: ask five new signups, "Did the demo match what you got after signup?"
Decision rules that keep you honest:
High play rate, low watch depth: a hook or length problem. Tighten the first 15 seconds.
High watch depth, low signup: CTA mismatch, pricing shock or friction after the click. Not a "make a cooler video" problem.
Low play rate: thumbnail or placement. Move it above the fold before you re-film.
Signups say "the demo showed X, the product is Y": re-record against the real success screen. Don't add more features to the video to paper over the gap.
Which traps waste a weekend?
Feature tourism: opening every tab "while you're here." Cold visitors came for one job.
The origin story tax: your founding arc doesn't belong in the first 20 seconds of a product demo.
Fake data that lies: demo accounts with impossible metrics train distrust. Use believable sample data.
Music louder than the voice: muted viewers get nothing from it and listeners get annoyed.
Captions as an afterthought: unreviewed auto captions that misspell your product name in the first line.
CTA drift: the video says "book a call," the page says "Start free." Pick one.
The endless re-cut: ship a B+ demo this week and improve it from the 14-day numbers.
What goes on a 90-second demo checklist?
□ One-job brief filled in (ICP, job, success screen, CTA, won't-show list)
□ Result shown in the first five seconds and again before the CTA
□ Script timed under 90 to 100 seconds at conversation speed
□ Clean demo environment and believable sample data
□ Mic test done; captions reviewed and not covering the UI
□ Muted, inline-friendly player with a visible play control if it autoplays
□ Thumbnail shows the win; player above the fold next to the CTA
□ CTA spoken in the video matches the button on the page
□ 14-day tracking: play rate, watch depth, watch to signup, five replies
□ One outsider watched a rough cut before you publish
Where should the demo go after launch week?
A converting demo is a distribution asset, not a vanity reel. Once it matches the product and the button on your page, put it where builders actually look: your landing page first, then launch threads, then directories that send cold traffic.
When you're ready to put that page in front of early users, explore what's launching on MakerHunt, and keep the video honest enough that the first ten minutes after signup feel like the same product they just watched.
After launch week, the demo keeps working wherever someone writes about your product, because a story that links to a page with a clear 90-second answer gives readers somewhere useful to land. Lede Press publishes product stories in a public daily edition, each written so a reader can understand what a product is and decide if it's for them. If you'd rather not pitch every month yourself, Posting Dude publishes 10 editorial articles about your product every month across a 10-site network (run by the same team as MakerHunt), with a Launch plan at $249/mo for one project and Studio at $399/mo for up to three. Neither replaces a page that converts; they are ways to put that page in front of more readers.
If your one-job demo keeps prompting "how do I do X?" questions, those are longer tutorials, not a longer demo. Start with starting a YouTube channel for your SaaS as a founder and keep the homepage cut at 90 seconds.
Frequently asked questions
Should a product demo video autoplay?
Only muted, and only if it makes sense without sound. Browsers allow muted autoplay but generally block sound until the visitor interacts. If you can't caption it well, use a click-to-play player with a strong poster frame instead.
Is 90 seconds right for a Product Hunt video?
Product Hunt's page-building guide doesn't set a length; it asks for a quick demo of the product doing its thing. The 90-second cut built for your landing page is a sensible default, since a gallery visitor is just as cold.
Should I build an interactive demo instead?
They do different jobs. Product Hunt lists a clickable walkthrough as a separate, optional media item so people can try the interface without signing up. Make the video first, because it also serves your homepage, and add a clickable demo later if visitors ask to try before signing up.