What it does, and what it does not

For an explainer: a topic and who it is for, in one box — or upload your material and say what the video should do with it. For a how-to: the website address and what the video should show. Everything else has a sensible default.

PDF, Word, PowerPoint, plain text and Markdown, up to 100 MB each. You can also add pictures for the scenes to use — a logo, a product shot, a diagram.

Choose Short (about a minute), Medium (two to three minutes) or Long (about five). The script is sized to fit, at a calm pace of about 150 spoken words a minute.

Four voices, recorded with ElevenLabs: two British and two American, female and male. They are multilingual, so the same voice can narrate in English, German and many other languages. You do not pick one: the studio chooses the voice from what you write, and you can ask for a man or a woman, or an accent, in your description. The studio itself is in English and German.

Yes. Turn on review and the build stops at the script — or, for a how-to, the storyboard. Edit it directly, or say what to change and get a rewritten version back. Nothing is recorded until you approve it. Leave review off and the whole video is made in one go.

Yes, by asking in the chat — “make scene 3 shorter”, for example. Only the scenes the change affects are made again; the rest stay exactly as they were.

It depends on the length and the kind of video. The build runs on our servers, so you can watch each stage in the workspace, or close the tab and come back to it later.

While exploring, it only looks around: it follows links and menus, scrolls, and reads what is on each page. It never clicks anything that deletes, removes, pays, buys, sends, publishes or signs out, and it never leaves the site. In the recording it performs the storyboard’s steps — clicking, typing, choosing, pointing — and still will not delete, pay, send or sign out unless you explicitly ask it to show that.

The details are encrypted before they are stored, used only inside the browser that records the site, and never sent to the AI: the storyboard uses placeholders and our own code types the real values. They are never shown again, not even to you. Where you can, use an account made for the recording rather than your own.

Not unless you ask. By default it signs in off camera and the recording starts on the first step. If signing in is the point of the video, ask for it to be shown — the username is then visible in the video.

The text of the pages the browser visits — headings, links, button labels and field names — so it can plan the steps. Never what is typed into a field.

No. The recorder only opens addresses on the public internet and refuses private network addresses, so it cannot be pointed at anything only our own servers can reach.