How to make a software tutorial video

01One task per video

The most common tutorial mistake is scope. A video called Getting Started that covers installation, configuration and first use is three videos, and someone who only needs the third one will not sit through the first two.

One task per video also fixes maintenance. When the settings screen changes, you re-record four minutes rather than twenty five, and the other videos stay accurate.

• Title the video with the task, in the words someone would search for
• State the starting point in the first ten seconds, so nobody is lost by step two
• Show the result before the steps when the result is not obvious
• Link to the prerequisite rather than covering it again

02Write the steps before you record

A tutorial script is not a word for word narration. It is the click path: every screen you will touch, in order, with the values you will type. Writing it down is what stops the recording where you realise at step seven that step three was wrong.

Read it once with the software in front of you. The gap between what the documentation says and what the interface actually does is usually found here rather than in the recording.

03Record in one take and pause between steps

Recording is controlled from the tray once the window hides itself, and paused segments are joined back together automatically, so there is no edit to make where the pause was. That changes how you record: instead of stopping when you need to think, read ahead or wait for something to load, you pause and resume.

The result is one continuous take with the waiting removed, which is far easier to produce than assembling four separate clips and matching their audio levels.

04Cut the dead air automatically

Automatic silence trimming detects and removes silent stretches from an audio track. On a long unscripted take this is the single biggest saving, both in editing time and in runtime: an eleven minute recording that becomes seven minutes without losing a word is normal.

Silence trimming and the six voice processing controls are Pro tier features. The free tier records and edits without them, which is enough to check the workflow on your own material first.

05Zoom where the click is

Automatic zoom follows the cursor, and for a tutorial that is usually correct because the cursor is where the instruction is. Intensity runs from 1.0x to 4.0x and the smoothing control decides whether the frame tracks instantly or lags behind the pointer.

The exception is any step where you are describing something the pointer is not on: a validation message, a value in a table, a state that changed elsewhere on screen. Those zooms can each be switched off cursor-following and pointed at a fixed spot instead.

06Add word level captions

Captions on a tutorial are not only an accessibility measure, they are how people skim. Someone who already knows the first half scrubs until the words change, and word level timing is what makes that work.

Transcription runs on your machine with a choice of model sizes, so no audio is uploaded and there is no API key or per-minute credit. Progressive word highlight dims the words you have not said yet at an opacity you set. Transcription is a Pro tier feature.

07Keep the pointer legible and out of the way

Twelve pointer styles with adjustable size and shadow, plus a tilt option that leans the pointer into its direction of travel. For tutorials the useful setting is hide-when-idle: the pointer disappears whenever it stops moving, so it is not sitting over the field you are asking someone to read.

Because the cursor is redrawn rather than captured, all of this is decided after the recording exists. You do not have to think about pointer size while you are trying to remember step nine.

08Export once, at the size the docs need

For an embed in documentation, 1080p at 30 fps is usually right and keeps the file small enough to load beside the page. Interfaces with dense small text are the case for 4K, where the Pro tier exports at 24, 30, 60 or 120 fps with GPU accelerated rendering.

GIF export deserves a mention for tutorials specifically. A single step that takes four seconds to show is often better as a loop in the documentation than as a video someone has to start.