Skip to content
AI Santa Video

Inside the workshop

How it works — drawn in your browser

Why a page about how it works

Most sites that promise you a video won't tell you how the video gets made. This one will, because the honest answer is more interesting than the marketing one: nothing is generated by a model, nothing is uploaded, and the whole thing runs on the device in your hands.

The chain, from keystroke to file

1. You type. A name and one line of wishes go into two input fields. They are held in the page's memory. No request leaves the browser.

2. The scene is drawn. A canvas element, 720 pixels wide and 1280 tall, is painted frame by frame in JavaScript: a night sky gradient, a workshop silhouette, a few hundred snowflakes each carrying its own position and speed, an envelope path, a wax seal, a waving figure built out of simple shapes and a warm palette.

3. Your words are set. The name is drawn with the browser's own text renderer, one character at a time, so it appears to be written. Your wishes follow on the next lines. Long lines wrap; very long words get clipped rather than spilling off the page.

4. The frames become a file. The canvas is asked for a media stream, and the browser's own recorder turns that stream into a video file in real time — twelve seconds of animation becomes twelve seconds of recording. When it stops, the file is handed to a download link.

5. The tab forgets. Close the page and the name, the wishes, the frames and the file are gone. There is no database row to delete, because no row was created.

What that means in practice

  • It works offline once the page has loaded. Try it on a plane.
  • It cannot leak your child's name, because there is nowhere for the name to go.
  • It has no queue. Every other visitor does not slow you down.
  • It is the same every time. Same name, same wishes, same scene, same video.

Where it can go wrong

The honest limitations:

  • Some phone browsers will not record a canvas. iOS Safari is the usual culprit. When recording is unavailable, the video still plays on the page and the site says so instead of handing you a broken download.
  • The file is WebM on most browsers, which every modern phone and desktop plays, but older Windows media players may not. Converting it is a five-second job in any online converter.
  • The text is drawn, not typeset. A name with unusual accents may render with the fallback font instead of the serif one.
  • Twelve seconds is twelve seconds. There is no length setting, because a longer video would mean a bigger file and a longer wait for a worse result.

Why not use an AI video model

Because the thing people actually want here is a name on a letter and a warm twelve seconds, not a photoreal video that takes minutes to render, costs credits, and has to upload a photo of a real person to work. A drawn animation is faster, private, free to run, and impossible to mistake for a deepfake.

If you want the photoreal version, there are excellent paid tools for it. If you want the one you can make in the next thirty seconds without giving anyone your data, you are on the right page.