Video teaser

Runs the bundled nanoodle workflow workflows/video-teaser.noodle-graph.json against the NanoGPT API: a one-liner (Tagline) feeds an LLM that writes a single cinematic shot description, which feeds a text-to-video model that renders a 5-second 480p clip (Teaser). Requires Node.js 22.23.4+, 24.22.1+, or 26.12.1+ and the nanoodle npm package (npx nanoodle fetches it).

API key

The run needs a NanoGPT API key. Prefer NANOGPT_API_KEY already in the environment. Otherwise pass --env-file <path-to-.env> (this CLI: --env-file overrides ambient env).

Never print the key.

Run

From this skill's directory (or prefix paths if running from elsewhere). Replace the Tagline value with the user's line:

npx nanoodle run workflows/video-teaser.noodle-graph.json \
  --input "Tagline=a slow cinematic drone shot over a misty mountain valley at sunrise" \
  --out ./teaser-out

Add --env-file .env only when the key is not already exported. Add --json for a structured result (paths, costUsd, remainingBalance). Video generation typically takes a few minutes — don't kill the run early.

Optional overrides (longer clips raise cost — see Cost below):

--input "Shot writer=<custom shot-writer instructions>"
--set "n3.duration=10"          # longer clip (roughly doubles the video cost)
--set "n3.aspect=9:16"          # portrait

For sharper video, edit n3.fields in the workflow JSON, keeping both resolution values in sync. For 720p (about $0.08/run), merge these fields:

{
  "resolution": "720p",
  "modelOpts": { "resolution": "720p" }
}

For 1080p (about $0.12/run), change both values to 1080p. The bundled graph sets both to 480p: the pinned 0.8.x executor uses fields.resolution, while the 0.11.0 CLI needs the explicit fields.modelOpts.resolution because it runs without a model catalog. --set "n3.resolution=720p" alone does not change the resolution sent by the 0.11.0 CLI.

Inspect the interface anytime with:

npx nanoodle inspect workflows/video-teaser.noodle-graph.json

Inputs

Shot writer is the node name shown in the editor. It replaces that node's system prompt.

Outputs

• Media is saved under --out as Teaser.<ext> — extension follows the video MIME (usually mp4). Use the path the CLI prints on the Teaser: line; do not hard-code the extension.
• With --json, the path is in outputs.Teaser.

Cost

Each run costs about $0.06 in NanoGPT credit at the default 5 seconds / 480p (ltx-2-19b bills per second by resolution; the LLM step is a fraction of a cent). Higher resolution or 10-second clips cost proportionally more. Runs spend real credit — tell the user before running repeatedly.

If a run times out or is interrupted after submission, the remote job may still be running. Check its status and retain successful outputs before deciding to submit again. A fresh CLI invocation is not a durable resume and may bill again; do not automatically retry the whole workflow.