Video Director

The Video Director writes, voices and renders a narrated motion-graphics video from one request. create_explainer turns a topic into an explainer, and create_launch_video turns a product description into a launch video. The director writes the voiceover and the shot plan, generates the voice, and aligns it word by word. Then it renders an MP4 in which every piece of text or graphic appears the moment the narration says it. It is built on the shot sequence pipeline, which you can also drive yourself.

The one-call tools, create_explainer and create_launch_video, exist only on Nodaro Cloud. start_video_director and the shot sequence tools work on every edition.

Where the Video Director is offered

A deployment offers or withholds the Video Director like a node. Its availability id is video-director, in the deployment's nodes.allow and nodes.deny lists. See Editions and profiles. Where it is not offered, create_explainer and create_launch_video are not listed, and POST /v1/video-director/run answers 403 node_not_available before anything is reserved. The run's speech, alignment and render steps are each checked as their own node: text-to-speech, forced-alignment and render-video. start_video_director stays listed. It only returns the guide, and every tool it drives is checked as its own node.

Motion graphics or animated footage

The Video Director makes motion graphics: kinetic typography, shapes and images, never illustrated or filmed scenes. For an explainer told through generated animated footage, use the video-explainer recipe instead; see Content recipes.

When you ask for "an explainer" without a style, the assistant asks you to choose before it calls any tool. It goes ahead when you have chosen, or when your request already named a style.

The three tools

The parameters are in the tools reference.

{ "topic": "How transformer self-attention works, in 60 seconds" }

{
  "brief": "Nodaro is an AI video studio. Paste a prompt, pick a style, get a finished video. For indie creators and small teams who don't have time to edit. Tone: confident, modern, no jargon."
}

Both tools return a job id. In clients with MCP Apps, such as Claude on the web, the card shows the progress and plays the finished video. Other clients poll get_job; the MP4 is at output_data.videoUrl. The video is also saved to your library.

What a video can contain

• Text, shapes and images appear in sync with the voiceover, each at the exact frame its phrase is spoken.
• Shapes are rectangles, circles and lines.
• Images can be freeform elements, or screenshots in the device-surface-showcase and cursor-ui-demo blueprints, which show your uploaded screenshots in a device or with a moving cursor.
• No live capture. The director cannot capture a website. create_launch_video with only a url answers that it needs a brief; with both, it ignores the URL. To capture a live page as screenshots, use the capture_site tool, where the install enables it.

What it costs

A one-call video runs these steps, each charged on its own:

• Refunds. The authoring credits come back if the run fails. Each later step is refunded only if that step itself fails: when the render fails, the voiceover, and the alignment if it ran, stay charged.
• An automatic repair. Sometimes the authored plan breaks a timing rule, such as two scenes that overlap. The director then gets one free corrective pass with the exact error, keeping the script and the cues unchanged. If that pass fails too, the job fails with a resolve: error.
• Doing it yourself. When the assistant writes the brief itself and runs the steps with start_video_director, there is no authoring charge: you pay only the voiceover, the alignment and the render.

Brand

Both tools accept a brand: a preset id from list_brand_presets, such as cobalt-corporate, or your own tokens with a palette, fonts and a logo. The palette and fonts apply to every scene.

{
  "topic": "How our new analytics dashboard works",
  "brand": {
    "palette": { "bg": "#0B0B12", "text": "#FFFFFF", "accent": "#8B5CF6" },
    "fonts": { "heading": "Montserrat", "body": "Inter" },
    "logo": {
      "name": "NODARO",
      "image": "https://…/uploads/logo.png",
      "imageBackdrop": "#111111"
    }
  }
}

Fonts can also carry a weight, a casing and a letter spacing for headings and body text separately. A value set on an element wins over the brand, and the brand wins over the blueprint's default. Letter spacing is never applied to Arabic text, because it breaks the joined letters. A weight that the font does not load falls back to the nearest one that it does.

Logo

With logo.image, the video always contains a logo-assemble-lockup scene, the only blueprint that shows the logo image. If the plan has none, the director adds it at the end. If the image fails to load at render time, that scene shows the animated wordmark instead, so a bad URL never fails the render. A render that starts right after the upload may show the wordmark before the image is available; render again after a short wait.

Author the video yourself

To review the brief before any render, ask the assistant to call start_video_director. The guide it returns covers choosing a genre (explainer or product-launch) and a story arc, writing the voiceover as cue phrases, and building the brief. Then the assistant runs the shot sequence steps: generate_speech, forced_alignment, resolve_shot_sequence and render_shot_sequence.

On that path, the brief takes brand tokens only as an object, not as a preset id: copy a preset's tokens from list_brand_presets. The logo scene is not added automatically, so include a logo-assemble-lockup reveal yourself.