On this page
The short version
- 1Pick one task, in the words people use to ask for it.
- 2Do the task yourself and write down every step.
- 3Turn the steps into a script that says why, not only where.
- 4Keep it to one to three minutes.
- 5Record it: with AI from a prompt, or by hand in short takes.
- 6Edit so people can follow along: zoom, pause, caption.
- 7Publish it with its steps and transcript, and re-check it after each release.
A tutorial isn't a demo
A demo persuades someone who hasn't decided yet. A tutorial helps someone who has, and is stuck. The differences change how you make it, so settle which one you need first. If it's a demo, read the product demo video guide instead.
| Tutorial | Product demo | |
|---|---|---|
| Who's watching | ||
| Who's watching | Someone using the product, mid-task | Someone deciding whether to buy |
| What it shows | ||
| What it shows | Every step of one task | The best three moments |
| Pace | ||
| Pace | Slow enough to follow along | As quick as it can be understood |
| It worked if | ||
| It worked if | They finished the task | They took the next step |
Pick one task
One task per video. “How to export invoices as a CSV” is a tutorial; “Billing overview” is a lecture. Small videos are easier to find, easier to follow and easier to replace when that one screen changes.
You already know which tasks to start with:
- Support tickets. The question you answer every week is your first video.
- Your help center's search box. It shows what people look for, in their own words.
- Where new users stall. The step before people drop out of onboarding deserves a video.
Short of ideas? In SayScreen, Discovery clicks around your app like a new customer and suggests videos worth making. Then title each video the way people ask: “How to make a QR code in Canva”, with the task first and the product last.
Write the steps first
Do the task yourself in a fresh account and write down every click, in order. You'll catch what experts forget: the setting that's on by default for you, the permission a new user doesn't have, the menu that only appears once a list is filtered. This one page is the whole plan.
- Title
- How to export invoices as a CSV in Taskly
- Who it's for
- Account admins closing the month
- Start page
- Billing, with last month's invoices already in the account
- Before you start
- You need the Admin role to see Billing
- Steps
- 1. Open Billing in the sidebar. 2. Filter by last month. 3. Click Export in the top-right corner. 4. Choose CSV. 5. Click Download.
- Watch out for
- Export only appears once the list is filtered
- Finished when
- The CSV is in your downloads, with one row per invoice
Write the script: say why, not only where
Every step needs three things: what to do, where it is, and what it's for. Use the button's exact label, so people can find it on their own screen. Warn them about the wrong turn before they take it.
| Instead of | Say | |
|---|---|---|
| Name the control | ||
| Name the control | “Now we click on this one here.” | “Click Export, in the top-right corner.” |
| Give the reason | ||
| Give the reason | “Next, fill in the form.” | “Give it a name your team will recognize. You can change it later.” |
| Warn before it happens | ||
| Warn before it happens | “Choose the second option.” | “You'll see Dynamic first, but skip it: Static has no Canva logo.” |
| Say what changed | ||
| Say what changed | (Silence after the click) | “A new tab opens with your report.” |
Budget 130 to 150 words a minute, and leave room for the screen: a page that takes three seconds to load needs three seconds of silence. The free script timer adds those pauses to its estimate, and the free online teleprompter scrolls the script at your pace when you record your own voice.
How long should a software tutorial video be?
One task, usually one to three minutes. The 4 tutorials in our own how-to library run from 66 to 90 seconds, and our tutorial video examples show what each one does with that time. Wistia's guide to video length puts how-tos in the 1 to 5 minute range. If a task runs longer, split it into parts that each end somewhere useful: connect the data source in one video, build the first report in the next.
Record it
How to make a tutorial video with AI
With an AI tutorial video maker, your plan becomes the prompt. In SayScreen, you paste the page where the lesson starts, sign in if it asks, on the site's own sign-in page shown inside SayScreen, and describe the task. The AI drives a real browser, reads each screen, clicks, and writes a line of narration for every step. A natural AI voice reads it, and you get an edited MP4 and an .srt file.
This is a real example from our library: the exact prompt, then the video it made.
Show how to make a free QR code in Canva. Open Apps in the left sidebar, search for QR code, and choose Static QR codes, not a dynamic one. If it isn't listed there, search for QR code in Elements instead. Enter https://sayscreen.com as the link and generate the code so it lands on the page. Mention that a static code is free, but its link can't be changed later. End by clicking an empty spot so the QR code shows on its own.
Check it against your app
Turn your plan into a video
$1 for your first 7 days, then $29 a month for 30 videos, up to 10 minutes each. Cancel anytime.
How to record a tutorial yourself
Record it yourself when your product isn't a web app, or you want your own voice in the room.
- Clear the stage. A demo account, notifications off, bookmarks hidden, and only the app in view.
- Make text readable. Use a window about 1280 pixels wide, so that people on small screens can read the labels you're pointing at.
- Record with what you have. The free online screen recorder runs in your browser, with nothing to install and no watermark.
- Take it one step at a time. Finish a step, pause, breathe, then start the next. Cutting the pauses out later is far easier than redoing a long take.
- Click slowly. Rest the cursor on the button and say its name before you click.
Edit so people can follow along
A viewer may be pausing the video and repeating each step on their own screen, so edit for following along rather than for pace.
- Leave a beat after each click, so that people can see what changed.
- Zoom in on small controls, keeping enough of the page in view to show where they are.
- Make clicks visible with a highlight, and show keyboard shortcuts on screen.
- Trim the waiting. Fast-forward slow page loads, but never cut out a step.
- Caption everything. Captions help people who can't turn the sound on, and non-native speakers.
SayScreen does most of this on every video: zooms that follow the action, a gliding cursor with click highlights, shortcuts as keycaps and trimmed waits. It writes an .srt file every time and burns captions in when you turn them on. By hand, you'll need a timeline editor such as Camtasia.
Publish it where people get stuck
Put the video on the page that answers the question: the help article, the onboarding email, the empty state in your app. Then give the people who'd rather skim a way to do that:
- Numbered steps under the video, each with a timestamp that jumps to its moment.
- A transcript, so people and search engines can read it.
- Captions in the format your player wants. Hosts usually take .srt and web players .vtt; the free subtitle converter switches between them.
- A short GIF for docs and changelogs, where a player won't fit. The free video to GIF converter makes one from a few seconds of the MP4.
- A “last checked” date, so readers know whether to trust it.
Our how-to pages are built this way, if you want a model to copy. SayScreen hands you the MP4 and the .srt; where you host them is up to you, and for YouTube it writes titles and a description and makes a thumbnail.
Keep it correct
Tutorials go out of date quietly: a button moves, a menu is renamed, and the video starts teaching something that isn't there. Three habits keep that small:
- Keep the prompt or script next to the video. Remaking it then takes minutes instead of a project.
- Check videos when you ship. When a release touches a screen, look at the tutorials that show it.
- Remake, don't patch. In SayScreen, run the same prompt again and check the result. By hand, re-record the steps that changed, and start over if the look of the whole interface did.
Mistakes that lose viewers
- A long intro before the first useful step
- Skipping a step because it's obvious to you
- Names on screen that don't match what you say
- Starting from a state new users won't have
- Text too small to read without full screen
- No end state, so people can't tell they're done
Questions
Give an AI tutorial video generator the page where the lesson starts and describe the task: the steps in order, the exact button names, anything worth warning about, and what the screen should show at the end. In SayScreen, the AI drives a real browser through your app, writes the narration and a natural AI voice reads it, and you get an edited MP4 and an .srt caption file. Check the steps against your app before you publish.
One task, usually one to three minutes. If the task takes longer, split it into parts that each end somewhere useful: connect the data source in one video, build the first report in the next.
A screen recorder, a microphone and an editor if you do it yourself: a browser-based recorder is enough to start, and desktop tools add zooms and timeline editing. An AI tool such as SayScreen replaces all three for web apps; desktop software still needs a recorder.
No. Viewers following a task want to see the screen, not you. You can narrate yourself, or use an AI voice. SayScreen narrates in a stock voice or your own cloned voice.
Use a demo or staging account with realistic sample data. It keeps customer information out of the video and lets you click anything without consequences. Whatever account you use, read every frame before you publish, including email addresses and account names.
Most tools create captions from the narration. Keep an .srt file for video hosts and a .vtt file for web players, and burn captions into the video for social feeds, where players often start muted. Read them through, because product names are what speech recognition gets wrong.
Check them whenever the part of the product they show changes, and put the date you last checked on the page. Keep each video's prompt or script next to it, so that remaking one is a small job rather than a project.