YouTube short and long script writer with web research

What it is

This n8n workflow writes a YouTube script from a topic. A form takes the topic and a choice between a Short and a long video. Tavily searches the web, OpenAI condenses what the pages say, and an agent writes the script. A Short gets a quick pass: three pages, one summary, one script. A long video gets ten pages, a summary of about 1,000 words and an outline before the script. Either way the script lands in a new Google Doc.

1. A form takes a topic and a Short or long choice
2. Tavily searches the web for the topic
3. OpenAI summarizes the pages, and outlines a long video
4. An agent writes the script in your sample style
5. Google Docs saves it as a new document

Who it suits

Creators who script their own videos, and the writers and agencies who script for others. It gives you a researched first draft to edit. Skip it if a script must be right on every fact, since its research is only the top pages one search returns.

Where to use it

Any n8n that can reach Tavily, OpenAI and Google will run it. The scripts end up in a Drive folder, where a team can comment on them.

• n8n Cloud: import the file into a new workflow there
• Your own n8n server: the same file and steps
• Google Drive: a folder of your choice holds every new document
• A team: share the form’s address so writers can request their own scripts

Setup

Get it saves a zip with workflow.json and Mindket’s setup notes in your downloads folder. Its author puts setup at about 13 minutes. Mindket replaced the two sets of sample scripts: in the original copy, each node held about 33,000 characters of its author’s own channel scripts.

1. In n8n, create a workflow, open its menu, choose Import from File and pick workflow.json.
2. Add an HTTP Header Auth credential with the name Authorization and the value Bearer, a space and your Tavily key. Select it in Tavily and Tavily1.
3. In those two nodes, replace <YOUR_TAVILY_API_KEY> in the Authorization header too, so both agree. The brackets go as well.
4. Add an OpenAI credential and select it in gpt-4o-mini, gpt-4o-mini1, WebSearch Summary and WebSearch Summary (Short).
5. Add a Google Docs OAuth2 credential and select it in Create Doc, Update Doc, Create Doc1 and Update Doc1.
6. In Create Doc and Create Doc1, replace YOUR_FOLDER_ID with your folder’s ID, the part of its address after /folders/.
7. In Reference Script, replace the placeholder with one or two of your Shorts scripts. In Reference Script1, put a long script.
8. Save, send the form once from its test address, then press Publish.

How to use it

Open the form from On form submission, type the topic in Provide topic and choose Short video or Long video. Both fields are required. The form confirms at once. The research and writing go on behind it, and the document shows up in your folder when the last node finishes.

Example

You type best free video editors for beginners and choose Long video. Tavily returns ten pages, OpenAI condenses them, and the outline lists each editor with its sub-points. The script follows that order and opens with a short hook. A Doc with the topic as its title appears in your folder. Choosing Short video instead gives a shorter script in a Doc titled best free video editors for beginners_Shorts.

For newcomers and experts

If you are new. n8n draws a workflow as boxes joined by lines, and each box, a node, does one job. This file has 19, but two are sticky notes that label the Short and long stretches. A credential is a key n8n stores for you and hands to the node you select it in. The style samples matter. The agent copies their tone and pacing.

For experts. Two OpenAI models share the work across four nodes, and the labels on the model nodes do not match what they run.

The prompts expect a listicle, a how-to or an experiment, but the form sends only Short or Long. The agent has to work the type out from the topic and the outline.

Good to know

Every run searches and calls the model several times, and the long path costs most.

• Billing: Tavily and OpenAI each bill their part of a run, and the long path sends up to ten pages of raw text to the model, which can be long
• Open form: once the workflow is published, anyone with the form’s address can start a run on your keys
• Check the facts: the script is the model’s reading of a few pages, so verify figures and names before you record
• The alias: chatgpt-4o-latest may be retired by OpenAI, so pick a current model in those nodes if a run fails there
• The notes: the original copy’s agent nodes carried a note naming its author’s channel. Mindket removed both
• Credentials: the JSON holds none, and nothing from its author’s accounts