Product Hunt We are on Product Hunt today, an upvote helps →

How whiteboard animation works

Not a video model. Four ordinary steps in the right order.

Updated: 2026-09-10

Step one: the script decides everything downstream

The narration is written or pasted first, because its length determines how many boards there are, and its wording determines what each board must show. Roughly twelve to sixteen seconds of speech per board is the pacing that reads as neither choppy nor stalled.

Step two: one board, one picture

Each board is generated as a single complete illustration with an explicit layout — left half, right half, quadrants — and the script segments are bound to those regions. Generating the whole board at once is what keeps the composition coherent; assembling separate cut-outs is what makes older tools look like collage.

Step three: finding the ink

The finished illustration is thresholded to separate ink from paper, then thinned to a one-pixel skeleton, then traced into stroke paths. This is why the drawing must keep large areas of paper visible: a dark background has no ink to find, and renders as noise.

Step four: timing the reveal

Speech synthesis returns a timestamp for every word. Each board segment starts revealing its region at the moment the voice reaches its first word, and later regions are subtracted from earlier ones so nothing is drawn twice.

Frequently asked questions

Is this generated by a video model?

No. It is a storyboard, still illustrations, speech synthesis and a stroke-order renderer, composed in that order.

Why does the drawing need white space?

Ink extraction looks for dark lines on a light page. Solid dark fills leave nothing to trace.

Why not animate every element separately?

Because composition falls apart. Generating the whole board keeps the picture designed rather than assembled.

How is the drawing kept in sync over four minutes?

Timing comes from word-level timestamps rather than a fixed frame rate, so drift does not accumulate.