Producing learning videos: planning, recording, editing

Learning videos are not one genre. Depending on the objective, different formats are suitable, and each has its own demands on effort and technology.

Every learning video begins with the question of what viewers should be able to do afterwards. That determines the length. For explainer videos, a few minutes have proven effective; longer content can be split into several videos that can be watched individually. That way learners can later find exactly the part they need.

Write a narration script or storyboard before recording. A freely spoken talk feels natural but almost always ends up longer and less precise. Read the script aloud before recording: what looks good on paper often sounds stilted when spoken.

Viewers forgive a mediocre picture but not poor sound. Echo, noise and fluctuating volume are tiring and make people switch off. An external microphone close to the mouth and a room with carpets, curtains or shelves improve quality more than any expensive camera.

For the picture, simple rules apply: calm background, even light from the front, camera at eye level. For screen recordings, a tidy desktop without private notifications and a font size that is readable on a phone are enough.

In editing, remove slips, pauses and digressions. Keep the pace calm; fast cuts tend to feel hectic rather than modern in a learning context. Background music can give structure but should stay quiet and drop further back during speech.

Captions are required for accessible videos and useful for everyone who learns without sound, for example in an open-plan office or while travelling. Automatically generated captions are a good start but should be checked for technical terms. Export the video in a widely supported format such as MP4 so that it plays in any learning platform.

An alternative to traditional production is generation from slides and narration. A synthetic voice reads the text, slides and graphics appear in time, and an avatar can appear on screen as the speaker. The great advantage is changeability: when a sentence changes, only that section is regenerated instead of reshooting.

Mind transparency: synthetic narrators and voices should be labelled as AI-generated, as the EU AI Act requires for certain content. And a real person may only be recreated as an avatar or voice with their explicit consent.

For many organisations a mix makes sense. Personal messages, such as a welcome from the management or a safety briefing at the real machine, are filmed traditionally. Explanatory content that changes often or is needed in several languages is generated from slides and narration. That way production effort goes where it makes the biggest difference.

In Universe, a learning video is created from the course: the presentation runs automatically, the speaker reads each slide and slides with pre-produced video play that video. You export one video per module or one continuous video with title, agenda and closing frame; recording runs in real time in the presentation view. Music moods duck during speech, and transitions apply in the player, the presentation and the video. Finished videos are never re-rendered automatically, only after approval.