How to Make a Whiteboard Explainer Video With AI in 2026
A working walkthrough — from a blank prompt to a rendered MP4 — including the prompt that produces a decent script on the first try and the three settings that matter.
Whiteboard explainers used to be a specialist job. Someone drew the frames, someone else timed them to a voiceover, and a two-minute video took a fortnight. In 2026 the whole pipeline runs from a paragraph of text. This is the actual walkthrough — what to type, what to change, and where the output usually goes wrong.
If you want the background on formats and when a whiteboard video is the right choice at all, start with What Is an Explainer Video? and come back.
Why whiteboard, specifically
Before the how, the one thing worth internalising, because it determines whether your video works:
The drawing is the retention mechanism. Not the voice, not the music, not the transitions. A line being drawn is an unfinished thought, and unfinished thoughts are hard to look away from. The moment your video stops drawing and starts displaying, you are relying on the narration alone to hold someone — and narration alone loses to a notification.
This has a practical consequence for how you write. Every sentence in your script should correspond to something appearing on the board. If a sentence has nothing to draw, cut the sentence.
Step 1 — Write the takeaway first
One sentence. The thing you want repeated to a colleague tomorrow.
"A CDN keeps a copy of your site near the user, so the page loads before the request would have reached your server."
Everything else in the script is scaffolding for that line. If you cannot write it, you are not ready to make a video — you are still figuring out what you think.
Step 2 — Write the prompt
Most people under-specify, get a generic script, and conclude the tool is bad. A good prompt names four things: the audience, the takeaway, the length, and what to leave out.
Weak:
Make a video about CDNs
Strong:
Explain how a CDN speeds up a website, for a non-technical marketing manager evaluating hosting. 60 seconds. The takeaway is that a CDN stores copies near the user so the page loads before the request would have reached the origin server. Use a concrete analogy. Don't mention pricing tiers or vendor names.
The second prompt produces a usable script on the first attempt roughly every time. The difference is not prompt magic — it is that you did the thinking.
Two more clauses worth adding when they apply:
- "Assume the viewer has never heard of X." Stops the script leaning on jargon.
- "End on the single thing worth remembering." Stops it trailing off into a summary of itself.
Step 3 — Check the scene shapes, not just the words
This is the step people skip, and it is the difference between a video that reads as a slideshow and one that reads as an explanation.
A whiteboard explainer is built from scene types, and a good generator chooses each one from the shape of the content:
| The content is… | The scene should be |
|---|---|
| Ordered steps, or cause and effect | A process flow with arrows |
| One whole and the parts inside it | A diagram with a centre |
| Points of equal weight, no order | Bullets |
| Two real alternatives, or before/after | A side-by-side comparison |
| One number worth pausing on | A stat |
| The single line worth remembering | A quote |
When you review a generated storyboard, read the scene types down the page before you read a word of narration. If they are all bullets and process, the script has been flattened — almost any idea can be forced into a list or a sequence, which is exactly why a lazy pass produces nothing else. Ask for the change explicitly: "make scene 3 a comparison" or "turn scene 4 into a diagram."
The other thing to check: do elements accumulate? A scene should be one board that fills up as the point is made, not a slide that gets swapped out. If every scene shows one icon and then wipes, you will end up with a video where nothing is ever on screen long enough to relate to anything else.
Step 4 — Make it work with the sound off
Assume nobody turns on audio. That is not pessimism; on social it is the majority case, and in an office it is nearly universal.
The test: read only the on-screen labels, in order, as a list. Do they tell the story?
Good:
Requests hit the origin → Origin is far away → Edge cache stores a copy → Page loads instantly
Bad:
The problem → It gets worse → Our solution → The result
The second set is not labels, it is a table of contents. Fix labels before you fix anything else — they are the only text most viewers will actually read.
Step 5 — Set the three things that matter
Everything else is a preference. These three change the output:
Aspect ratio. Portrait 9:16 for Shorts, Reels and TikTok. Landscape 16:9 for your website, an LMS or a sales email. Square 1:1 for feeds. Pick it before you review the layout, because the layout genuinely differs — a six-item board that is comfortable in landscape is cramped in portrait.
Length. 60 seconds is the workhorse. Below 30 you cannot establish a mechanism; above 120 you have two videos. Note that length also changes density: a longer video does not just add scenes, it gives each board more room to fill up.
Voice. Listen to two or three on your actual script rather than on a demo sentence. Pace matters more than timbre — a voice that is 10% too fast makes a technical script unfollowable, and you will not notice on a one-line sample.
Captions are worth leaving on for social and worth turning off for anything embedded in a page next to written text, where they just duplicate what is already there.
Step 6 — Preview, revise, then render
The most useful property of an AI whiteboard tool is not speed, it is that revisions are cheap. Use that.
Watch the preview once without reading along. The failures you are looking for:
- A drawing that arrives before or after its sentence. Means the timing is estimated rather than measured. Good tools transcribe the finished audio to get word-level timing; if yours drifts, that is why.
- A scene that is one icon and a wipe. Nothing accumulated.
- The same icon twice in one scene. Kills a comparison instantly.
- A generic opening image. If your video opens on a lightbulb or a play button, it opens exactly like every other video. The first thing drawn should be the object the video is about — a stethoscope for a clinic workflow, a shopping cart for checkout.
Then ask for changes in plain English: "make the third scene a comparison of managed vs self-hosted", "shorten the opening to one sentence", "use a different icon for the second step". A good tool only rebuilds the scenes you touched, so a revision takes seconds rather than regenerating the whole video.
Render last. Rendering is the only step that costs anything.
What AI still gets wrong
Worth knowing so you can catch it:
- It over-explains the setup and under-explains the mechanism. The middle is where the teaching happens and it is consistently the thinnest part of a generated script. Push on it.
- It reaches for abstractions in the visuals. "Synergy" and "optimisation" are not drawable. Concrete nouns are: a padlock, an hourglass, a rocket, a database. If the concept is abstract, the drawing will be generic.
- It writes for the eye, not the ear. Watch for parentheses, semicolons and abbreviations that no narrator would say aloud.
- It defaults to one rhythm. Left alone, every video it makes has the same structure. Varying the scene types is how you avoid a library of videos that all feel like the same video.
A realistic time budget
| Step | Time |
|---|---|
| Writing the takeaway and the prompt | 10 min |
| Generating the storyboard | ~10 sec |
| Reviewing scene shapes and labels | 10 min |
| One or two revisions | 2 min |
| Rendering the MP4 | ~1 min |
About 25 minutes, and 20 of them are you thinking. That ratio is the actual change in 2026: the production time went to nearly zero and the thinking time stayed exactly where it was.
Frequently asked questions
Can I make a whiteboard animation for free? Yes. Vidmoro plays the whole video in your browser before you render, so writing, watching and re-rolling costs nothing — you only spend on the final MP4.
Do I need to be able to draw? No. The drawings are generated from the concepts in your script and animated stroke by stroke.
How long should a whiteboard explainer be? 30 seconds for one idea, 60 for a mechanism, 90–120 for a process with real steps. Past two minutes, split it.
Can I use my brand colours? Board styles cover the common looks — classic marker, chalkboard and blueprint, each with its own stroke weight and accent colour.
Is a whiteboard video better than motion graphics? For teaching, usually yes, because the drawing paces itself to the narration and holds attention while the picture is unfinished. For brand-led launches, motion graphics.
What is the difference between this and a faceless video? Different job entirely — one explains, one earns attention in a feed. See Explainer Video vs Faceless Video.