How to write a training video script: a situation, one idea per sentence, sources

The script decides everything. Good animation cannot save text written to be read, and a good script holds up with simple drawings. This guide draws on the videos generated with EducPilot, where every word, duration and source is known.

Before you write: three questions

1. Who for? The viewer, what they already know, and where they will watch (a desk, a phone, a meeting room).
2. To do what? What they will do differently afterwards. One sentence, one verb.
3. How long? Your word budget follows from it.

Across our own videos, narration runs at 95 to 130 words per minute, pauses and drawing time included. A 45-second video is about 85 words; a two-minute video, about 250. That is not much: you have to choose.

The five-part structure

This comes from the video AI at work: 5 rules, which runs 1 min 44 s.

Why open on a situation

A definition up front ("CEO fraud is a type of…") asks for effort before giving a reason to make it. A situation gives the reason first: a time, a place, an action the viewer has already taken. Three openings from our videos:

• "Friday, 5:55 pm. Your CEO calls."
• "That PDF you emailed? It is not an e-invoice."
• "Still typing the same password on every site?"

One idea per sentence

In our videos, a sentence averages six to eight words. Each carries one idea, and a sentence is also the unit of correction: change one, and only that one is re-recorded.

Before:

It should be noted that, within the framework of the e-invoicing reform, all companies, regardless of their size, are required to be able to receive electronic invoices.

After:

Since September 2026, every company receives its invoices through an approved platform.

Same information, half the words, and a concrete subject.

Write for the ear

A script is spoken, not read. A few rules:

• Numbers as you say them: "September first, 2026", not "09/01/2026".
• No brackets, no footnotes: the ear cannot hear them.
• Acronyms explained once, or avoided.
• "You" and "we", rather than impersonal phrasing. Mayer calls this the personalization principle: a conversational tone helps people learn.
• Repetition is allowed in the recap: that is what fixes the keywords.

Read the script aloud with a stopwatch. Wherever you stumble, viewers drop off.

A source for every fact

A training video states rules. Every fact must trace back to a dated document:

• note the source and page next to every sentence that claims something;
• prefer the official source to an article summarising it;
• date the video, and check again when the text changes.

In the e-invoicing video, each claim links to its page in the tax administration's PDFs: page 5 of the FAQ for "a plain PDF is not an e-invoice", page 6 for the timeline. That is what makes review fast: the expert checks quotes, not impressions.

What the picture does meanwhile

The script tells, the picture shows. It does not copy the sentence:

• one drawing and one keyword per idea;
• a comparison when the script contrasts two things (a weak prompt and a strong one);
• a diagram when it describes a sequence (the three steps of a sign-in);
• subtitles separately, for people watching without sound.

The guide Whiteboard explainer videos for corporate training covers these devices.

A full example: 58 seconds, 101 words

The script of Passkeys in 60 seconds, in full, with the job of each line:

Reviewing the script

• Read aloud, timed
• One idea per sentence, no sentence over 15 words
• Every fact tied to its source
• Reviewed by the subject expert
• Reviewed by someone from the target audience: do they get it without help?

With EducPilot

The assistant writes a first script from your documents along this structure, and submits it as a plan before producing anything. You then correct on the video itself: a note pinned at the right moment, or a sentence edited directly. Only the changed sentences are re-recorded. For the whole method, from document to published video, see How to turn a PDF into a training video.

Sources

• Richard E. Mayer, Multimedia Learning, 3rd edition, Cambridge University Press, 2020 (coherence, segmenting and personalization principles).
• Philip J. Guo, Juho Kim, Rob Rubin, How video production affects student engagement: an empirical study of MOOC videos, ACM Learning at Scale, 2014: short videos, with instructors speaking fairly fast and with enthusiasm, hold attention better.
• Durations and sentence lengths are measured on EducPilot’s five example videos (the WebVTT subtitles published with each video).