Quick start

From "I just signed in" to "the pipeline is rendering my video" in about three minutes. The video itself takes longer, around fifteen to twenty-five minutes start to finish, depending on length and format, but the part you have to be at the keyboard for is short.

What you need before you start

A YouTube channel you actually want to publish to, and one line that describes the video you want. That's the whole pre-flight checklist.

You don't need a script. You don't need a thumbnail. You don't need stock footage. You don't need to pick a voice in advance. The pipeline handles all of that and surfaces the choices when it's time to make them.

The five-minute flow

Click New video on the dashboard. The brief form opens with a single big text field at the top.

Write one to three sentences about the video. The shape that works is: what's the topic, what angle do you want, what tone. Something like "The Bletchley Park codebreakers and how Turing's bombe machine actually worked. Sober documentary tone, 12 minutes, primary-source quotes where you can find them."

Pick the format. Documentary is the default, and it's what most long-form history / commentary / explainer videos want. The format chips are right under the brief.

Pick the length. Four options: 8, 12, 20, or 30 minutes. The pipeline will write the script to fit, give or take a few seconds.

Pick the niche. CueTheScene has thirty-one niches with their own footage routing, voice biases, and prompting fragments. History, true crime, science, finance, gaming, tutorials, the lot. The niche picker has a search; type a word and the matches narrow.

Click Submit. The pipeline takes it from there.

What happens next

The job moves through six stages. You can leave the tab open and watch, or close it and come back when the email lands.

1. Outline. Claude writes a scene-by-scene outline. Twenty to forty seconds.
2. Script. Each scene gets a script that matches your length target. About a minute per minute of video.
3. Voice. ElevenLabs synthesises the narration at the voice you picked. A few minutes for a longer video.
4. Footage. The footage orchestrator pulls clips from the niche-appropriate sources (archival for history, Pexels and Pixabay for stock, the curated CueTheScene library for hero shots). A vision model rates each candidate against the scene script and picks the best.
5. Render. Shotstack assembles the timeline. Three to eight minutes for a typical 12-minute video.
6. Publish. Either you publish to YouTube from the dashboard, or you download the MP4. Your call.

You'll get an email when each major stage lands, and the dashboard live-updates without a refresh.

When to step in

You don't have to step in at all. The "auto" path is the default and it's the one most creators take most of the time.

If you do want to step in, the natural moments are:

• After the outline lands, open it, read it, regenerate sections that miss the angle you wanted. This is the cheapest place to redirect the video.
• After the footage lands, the editor lets you swap any clip the AI picked. There's a per-scene "Search more" affordance that pulls fresh candidates from any of the source tiers (All, Archival, Free stock, Pixabay).
• Before publishing, the title, description, tags, and thumbnail are all editable on the publish surface. The AI's first picks are usually good; the editor lets you make them yours.

When something looks wrong

The first move is to read the scene that looks off and check what the script said vs what the footage showed. Nine times out of ten the script asked for something the available footage couldn't quite deliver, and the fix is to regenerate just that scene, or to use Search more and pick a clip yourself.

The dashboard surfaces flagged scenes proactively, the editor's confidence badges show which scenes the vision model wasn't sure about. Start there.

What's next

Your first video is the long version of this page, with every choice the brief form offers, when each one matters, and the prompting tricks that lift script quality the most.

Editor overview covers the post-generation editor, what each panel does, the keyboard shortcuts, the regenerate-with-hint flow.

If you want to see the prices first, Plans and pricing has the full breakdown.

Cheers,
Carl