Questions people ask first

No. Your video file stays on your computer in every setup, and pictures are cut from it there. What else leaves depends on the model you choose for each job. A model running on your own computer sends nothing at all. A cloud service is sent only what that job needs: the sound of your recording to turn it into a transcript, or the transcript text to write the draft. Settings lists every model you have set up and says what each one sends.

Several gigabytes and an hour or more is the normal case, not the limit. Long files are the design target, so nothing loads the whole video into memory and progress is reported honestly from the first second.

You nudge it. Timestamps get you close, and the frame adjuster is how you land it exactly. Because images store their timestamp rather than a flattened bitmap, you can re-pick a frame a week later without touching the rest of the article.

It sounds like your walkthrough, tightened. The draft follows what you actually said and in what order. Then you edit it in the finished layout, which is where most of the voice comes from anyway.

Yes to both, and every feature works on all three paths: our built-in AI service, your own API key, or models running on your own computer. Bringing your own key means a console key from that provider with billing enabled on it, held on your machine and used directly, so the work goes to your account rather than through us. Choosing models on your own computer means the application needs no network at all.

Markdown with a folder of pictures, HTML with HowTo structured data included, or straight into WordPress or Webflow. The export is the same document you were editing, so nothing reflows on the way out.

The export is a bundle built to be read by an AI agent rather than by a person copying and pasting. It carries a manifest as its index, the tutorial as a structured document tree with no markdown to parse, the same tutorial as markdown, the pictures, and the typefaces the text on them is drawn in. Point your own agent or script at that folder and it has everything it needs to lay the guide out in your help centre, your CMS, or your docs site, whatever those happen to be. If you would rather paste one file in by hand, the HTML export is a single page with the structured data already in it.

$10 a month, which covers the application on up to 5 machines. If you use the built-in AI service, you buy credit for that separately. If you would rather link your own AI service with an API key, or run a model on your own computer, both are supported in the application — and on either of those the $10 is your whole cost. To put a real figure on the built-in service: a 14 minute recording cost ~$0.15 to extract the transcript from and ~$0.53 to turn into a 1,324 word tutorial, so ~$0.68 for the finished article. That is a job measured in September 2026, it is what you would have paid rather than what it cost us, and these rates follow what our own providers charge us, so they can change.

Credit pays for the built-in AI service, and only for that. It is spent on the two jobs that need a model: extracting the transcript from your recording, and writing the draft from that transcript. Those go to different providers, because the best model for listening is not the best model for writing. Bring your own API key or run models on your own computer instead and no credit is used at all. You are charged in dollars and cents rather than in tokens or points, and each job shows an estimate before it runs and the real figure after. No two runs cost quite the same: a longer recording costs more to listen to, and a guide the model has to think harder about costs more to write. Credit never expires, and it is not refundable.

Up to 5 at once. You can see them all in your account and revoke any of them, which frees the slot immediately. A revoked machine can activate again later if a slot is free.