Top 10 countdown videos made and uploaded to YouTube daily

Top 10 countdown videos made and uploaded to YouTube daily

An n8n workflow that turns a queue of subjects in Google Sheets into one top ten countdown video a day. Models on
OpenRouter fill the queue with ideas, then write an intro, an outro and ten ranked entries for the next subject.
JSON2Video renders the video from a template, with an AI voice and a picture for every scene, and YouTube gets it as an
unlisted upload. Nothing goes public by itself.

What is inside

• workflow.json: the workflow as n8n exports it, 31 nodes. Six are sticky notes that only name the stages. It holds no
  credential.
• README.md: these notes.

What we changed from the original copy

• The author's sheet and its tab id are gone. Google Sheets, Add to Sheet, Created Video, Error and
  Update Sheet ask for YOUR_SHEET_ID and find the tab by the name Sheet1.
• A setup note with the author's name, his channel, a link to his copy of the sheet and a referral link is gone. Its
  steps are below.
• The OpenRouter, JSON2Video, Google Sheets and YouTube credentials are gone, along with the author's account names on
  them.
• Nothing else changed. The JSON2Video template ID, the voice and the image model are as the author left them.

Before you start

• An n8n instance, on n8n Cloud or your own server.
• An OpenRouter API key with credit.
• A JSON2Video account and its API key. Renders on the free plan carry a watermark.
• A YouTube channel, and a Google account for the YouTube and Google Sheets credentials.
• An ElevenLabs API key saved in JSON2Video, or the switch in step 6 to JSON2Video's own key.

Set it up

1. In n8n, create a workflow, open its menu and choose Import from File. Pick workflow.json.
2. Make a Google Sheet whose first tab is named Sheet1, with the four headers below in row 1. Copy its ID: the long
   part of its address between /d/ and /edit.
3. Add a Google Sheets OAuth2 credential. Select it in the five sheet nodes and replace YOUR_SHEET_ID in each with the
   ID.
4. Add an OpenRouter credential with your key and select it in 4o-mini and 4o.
5. Add an HTTP Header Auth credential with the header name x-api-key and your JSON2Video key as the value. Select it
   in Generate Video and Get Video.
6. Choose the key behind the voice. Generate Video sends voice.ConnectionID as my-elevenlabs-connection. In
   JSON2Video, open Dashboard, then Connections, and save your ElevenLabs key under exactly that ID. Or change the value
   to none, and JSON2Video voices it with its own key.
7. Add a YouTube OAuth2 credential for your channel and select it in YouTube.
8. Save. Type one subject into the sheet with both statuses at To Do, then run the workflow from
   When clicking ‘Test workflow’. Press Publish once the row reads Created and Unlisted.

The sheet

Row 1 of Sheet1 holds these four headers, spelled as here:

| Column | What goes in it |
|---|---|
| Subject | A topic of four words or fewer, such as Board games |
| Creation Status | To Do until the video renders, then Created |
| Posting Status | To Do until the upload, then Unlisted |
| Video URL | The rendered file's JSON2Video address, or Error |

How a run goes

Both schedules fire every day at 10:00.

1. Schedule Trigger1 starts the idea branch. AI Agent asks 4o-mini for ten subjects that do not depend on the
   news. Split Out and Loop Over Items append each one as a row, with both statuses at To Do.
2. Schedule Trigger, or the manual trigger, starts the video branch. Google Sheets reads the first row whose
   Creation Status is To Do. With no such row, the run ends there.
3. Intro & Outro asks 4o, which runs openai/gpt-4o, for a spoken intro and outro and an image prompt for each.
4. Ranker asks the same model for ten entries, from number 10 down to number 1. Each has a voice line of two or three
   sentences, an image prompt and a short label for the screen.
5. Generate Video posts all of it to JSON2Video's v2/movies as variables of the template named in its body,
   with an ElevenLabs voice and flux-pro images.
6. 150 Secs waits two and a half minutes. Get Video asks for the project's status and Switch routes it. On
   running or preparing, 10 Secs waits and the status is asked again.
7. On done, Created Video marks the row Created and writes the file's address. Download Video fetches the file,
   and YouTube uploads it as unlisted, titled Top Ten: and the subject. Update Sheet sets Posting Status to
   Unlisted.
8. On error, Error writes Error into Video URL and the run stops.

Change the style

• The subjects: the system message in AI Agent and the sample list in Structured Output Parser.
• The script's tone and length: the system messages in Intro & Outro and Ranker.
• The models: 4o sets openai/gpt-4o. 4o-mini sets none, see Good to know.
• The voice and the pictures: voiceID, voiceModel and imageModel in the body of Generate Video.
• The upload: title, description, category, region and privacy in YouTube. Category 1 is Film & Animation.
• The hour: each schedule trigger, in the time zone your n8n uses.

Good to know

• Retired models: OpenAI shuts down GPT-4o and GPT-4.1 mini on 14 January 2027, and names gpt-6-luna as the
  replacement. OpenRouter lists it as openai/gpt-6-luna, and still lists both older models today. Change the model in
  4o, and set one in 4o-mini, before that date.
• The idea model: 4o-mini names no model, so n8n uses the node's default. In current n8n that is
  openai/gpt-4.1-mini, whatever the node's name says.
• A queued render ends the run: JSON2Video reports pending while a render waits for a worker, and Switch has no
  rule for it. Its fourth rule tests preparing, a value missing from JSON2Video's list of states. Change that value to
  pending and a queued render is polled like a running one. As shipped, the run stops without an error and the
  subject comes up again the next day.
• Quotes break the body: the scripts go into the JSON of Generate Video between double quotes, with nothing escaped.
  A line the model writes with a double quote in it makes the body invalid, and the call fails. Wrapping each value in
  JSON.stringify() and dropping its outer quotes avoids that.
• Failed renders repeat: Error leaves Creation Status at To Do, so the same subject is first in line again the
  next day. Change its status by hand to skip it.
• The queue grows: the idea branch adds ten subjects a day and the video branch uses one. Set Schedule Trigger1 to
  weekly, or turn it off once the sheet holds enough.
• No polling cap: a render that stays at running is asked about every ten seconds until it changes. JSON2Video says
  to treat a render still running after 15 minutes as failed.
• A short-lived file: JSON2Video deletes a rendered video after 7 days, so the address in Video URL stops working
  then. The YouTube link is never written to the sheet.
• The template lives in JSON2Video, not in the file. If JSON2Video answers that it cannot find it, copy a template that
  reads the same variables into your account and put its ID in template.
• The voice key: a connection ID JSON2Video does not know fails the render. Save the key under
  my-elevenlabs-connection or send none.
• Costs: OpenRouter bills three model calls for each video and one for each batch of ideas. JSON2Video bills one credit
  for each second of video. The original copy's setup note put a flux-pro image at 20 credits, against none for
  flux-schnell, and a video asks for twelve images. Check the current price in your JSON2Video dashboard before the
  schedule runs daily.
• Your data: the subjects and the scripts go to OpenRouter and JSON2Video, and the video goes to YouTube. The sheet
  stays in your Google account.
• Credentials: the JSON holds none, and nothing from its author's sheet.