Skip to content

Making teaching videos ​

A good teaching video changes what someone can do afterwards. This guide turns research on how adults learn into a repeatable plan for a lesson in Formercial. It suits doodles best, but it works with any kind of scene.

The lesson pattern ​

Plan each video around one concept, with these parts in order:

PartWhat the viewer getsWhy
HookA real question or puzzleGives them a problem to solve
Why it mattersThe life or work problem this solvesAdults want to know why before they commit their time
Try it firstA small task, before any explanationTrying first prepares them for the explanation
Reflect"What did you notice?"Connects the task to what they already know
The ideaThe explanation, in short sectionsThe rule makes sense because they've met the problem
ChallengeA new problem, in their own context, with a cue to pauseThey apply the idea without your help, before they see the answer
CheckThe answer, or a way to verify itThey can correct mistakes while the problem is still fresh

The pattern sets the order, not the words. If every video opens with "Why does this matter to you?" and closes with "Here's your challenge", it starts to sound like a template, and viewers tune out. Keep what each part does, but say it differently each time: "So why draw at all?" does the same job, and sounds like a person talking.

Build it in Formercial ​

  1. Start from a lesson preset. Each one teaches one thing using this pattern, and its stage scenes are named hook, why, try, reflect, challenge and check, so you can see what to replace. Each preset words these moments differently, to show that it can be done:

    bash
    pnpm kick formercial new "My lesson" --preset doodle-light-bulb

    For a code lesson, use javascript-const-let or typescript-types. To draw the ideas and then show the code, use python-objects.

  2. Write one idea per scene. Put the narration in spoken. Short scenes are easy to reorder, cut and reuse.

  3. Generate the voice before you time the drawings. Then set at on each stroke, so it appears as the matching word is spoken. See Doodle videos.

    bash
    pnpm kick formercial voice my-lesson
  4. Leave real time for pause moments. A scene lasts at least as long as its voice, or as long as minSeconds if that's longer. So on a "try it first" or challenge scene, ask viewers to pause, and set minSeconds to a few seconds more than the clip. Captions end with the voice, so put the prompt in a block to keep it on screen:

    json
    {
      "id": "try-it",
      "kind": "blocks",
      "minSeconds": 20,
      "caption": "Pause the video, and write down your guess.",
      "spoken": "Pause the video, and write down your guess.",
      "blocks": [{ "type": "headline", "text": "Write down your guess." }]
    }
  5. Keep captions available. Every render also writes an .srt file, which you can upload to YouTube. With captions on, it's in renders/srt/. If burned-in captions cover your drawing, set "captions": false, and the file sits beside the MP4 instead.

  6. Add chapters when you publish. Put timestamps in the video description, starting at 00:00, so viewers can jump to what they need. The .srt file shows when each line starts.

  7. Render. Long lessons take a while, so check the first minute before you render it all:

    bash
    pnpm kick formercial render my-lesson --format 16:9 --seconds 60
    pnpm kick formercial render my-lesson --format 16:9

Checklist ​

  • [ ] The video teaches one concept, and you can name it in one sentence.
  • [ ] The opening names a real problem the viewer has.
  • [ ] Viewers try something before you explain it.
  • [ ] New ideas connect to something they already know.
  • [ ] The challenge is a new problem, not a repeat of your example.
  • [ ] Viewers are told to pause before the answer appears.
  • [ ] Viewers can check their answer straight away.
  • [ ] The recurring moments don't reuse the wording of your other videos.
  • [ ] Chapters let experienced viewers skip ahead.
  • [ ] Every label is readable on a phone.
  • [ ] Captions are burned in or uploaded.

What the research says ​

Treat these ideas as design guidance, not laws. The strength of the evidence varies a lot.

IdeaWhat it saysHow strong it is
Andragogy (Knowles)Adults want to know why, direct themselves, draw on experience, learn when it's relevant, focus on problems, and are motivated from withinA useful guide, but critics point to limited empirical evidence and overgeneralisation. Younger learners benefit from many of the same things, and it may not fit every culture.
Experiential learning cycle (Kolb)Learn by experiencing, reflecting, forming a rule and testing itA sensible way to structure a lesson. Don't sort viewers into Kolb's "learner types": a major review found no adequate evidence for teaching to learning styles.
Short videosIn a study of 862 edX videos, engagement dropped sharply after about 6 minutesStrong for watch time, but that measures engagement, not learning.
SegmentingSplitting material into segments improves learning, with small to medium effects on retention and transferSolid: a meta-analysis of 56 investigations. It doesn't set a best length.
Problem firstSolving a problem before instruction can beat instruction firstStrong, in a meta-analysis of 53 studies. Check how well it carries over to video, since a viewer can skip the attempt.
Retrieval practiceTesting yourself beats restudying on delayed tests, even though restudying feels more effectiveStrong and widely replicated.
Spaced practiceSpreading practice out helps retention, and the best gap depends on how long you need to rememberStrong, based on 317 experiments on verbal recall. Revisiting earlier ideas in later videos applies it.
FeedbackOne of the most powerful influences on learning, but it can help or harmStrong overall, but findings on immediate versus delayed feedback are mixed. Giving the answer in the video is a practical default, not a proven best.

A video can't watch how someone responds. So the pause, the challenge and the check matter most: they're what get viewers doing something rather than only watching.

Released under the Apache License 2.0.