Stop Staring at the Cursor: How to Build a “Video-First” Content Pipeline with Make and Notion

Author(s): Anna J

Originally published on Towards AI.

“Video-First” Content Pipeline

A practical guide to turning your raw videos into great SEO articles from scratch without writing a single word.

We have all been there. It is 9:00 in the morning. You have a fresh cup of coffee. You have a great idea for an article that can establish your authority in your field. You open a blank document. And then you become stable.

The blinking cursor makes fun of you. A sentence that seemed very smart in your mind five minutes ago now seems strange. You write a paragraph, delete it, check your email, and suddenly it’s afternoon. You have created zero content.

This is not a creativity problem; This is a friction problem.

Most professionals—whether you’re a project manager, consultant, or junior developer—are better speakers than writers. When you’re on a Zoom call with a client, the insights flow naturally. When you are explaining a concept to a junior colleague, you are clear and precise. But as soon as your fingers hit the keyboard, that flow state evaporates.

The solution isn’t to force yourself to become a better writer overnight. The solution is to change the input mechanism.

We need to stop writing from scratch and start “creating” content.

In this guide, we are going to build a Video-first content pipelineWe’ll build a “no-code” automation system where you simply drop a raw video file (a Loom, a Zoom clip, or a voice memo) into a Google Drive folder, and a few minutes later, a perfectly formatted, SEO-optimized blog post will appear in your Notion workspace,

We are not using AI to “generate thoughts” or hallucinate. We’re using AI to translate your real voice into text.

Here’s how the machine is built.

The “Lazy” Stack: What You Really Need

Before we start connecting wires, let’s look at the toolkit. We’re avoiding expensive enterprise software in favor of tools you already use, connected to a powerful automation hub.

Here’s the stack:

Google Drive: Trigger. It’s reliable, it handles large files well, and it syncs with your phone.
Make (formerly Integromat): Nervous system. We’re choosing Make over Zapier for this specific build because its visual interface allows for better data manipulation (especially JSON parsing, which we’ll get to later).
OpenAI API (Whisper + GPT-4o): Brain.

Comment: You need not only a ChatGPT Plus membership, but also a Payments API account loaded with credit.

perception: Destination. This is where your content calendar lives.

A note on cost: This workflow isn’t free, but it is cheap. It costs a lot of money to process a 10-minute video through the OpenAI API. Compare this to the hourly rate of staring at a blank screen for two hours, and the ROI is mathematical, not theoretical.

Step 1: Trigger Mechanism (Google Drive and Make)

The goal here is simplicity. You do not need to be logged in to the dashboard to start this workflow. You should be able to move a file by dragging it to a folder.

Configure Google Drive

Go to your Google Drive.
Create a new folder. Name it exactly: _INPUT_RAW_VIDEO,
Create a sub-folder named (optional) Processed Where we will transfer the files later so that the system does not play the same video twice.

Configure Make.com

Click on the big plus button and search Google Drive,
Select trigger: View files in a folder,
Connect your Google account and select your _INPUT_RAW_VIDEO folder.

“Hidden Gems” Settings: By default, this trigger may attempt to process Any The file you drop there, including a PDF or an image. This will break our automation.

Right-click on the Google Drive module.
set a filter,
Label it: Media Files Only,
Situation: MIME Type prevent audio Or MIME Type prevent video,

Now, if you accidentally drop a spreadsheet in that folder, the automation ignores it. It’s a small detail, but it prevents 90% of error messages.

Step 2: From Audio to Text (The Whisper Model)

Most people try to use standard dictation tools for this. He fails. Always. Standard dictation creates a wall of text without punctuation, and it usually crushes technical terms.

We are going to use OpenAI whisper Sample. It’s surprisingly good. It understands heavy accents, loud conversations and even ignores “ums” and “ahhs” to some extent.

Net

In Make, add a new module connected to your drive trigger: OpenAI,
Select action: Create a transcription (whisper),
In the “File” field, map the file that comes from the Google Drive module.

“Hidden Gems”: Reference Parameters

Here’s the secret that most tutorials miss. In the OpenAI module settings, there is a field called ready (optional).

Don’t leave it blank.

Whisper uses this sign to understand Context And spelling of your audio. If you work in a specific industry, this is essential.

For example, if I’m recording a complaint about automation, I’ll paste this into the prompt field:

The transcript is about no-code automation, SaaS, API integration, JSON, Make.com, Zapier, and Python scripts.

By feeding these keywords, Whisper knows exactly what I mean when I say something that sounds like “Jason.” JSONThis simple step saves you 20 minutes of editing later,

Step 3: Intelligence Layer (Context Injection)

Now we have a raw transcript. It’s accurate, but it’s messy. It feels like a verbal conversation – full of tangents and repetitions. We need to turn this into an article.

add another OpenAI module, but this time choose Create a closing (chat)We want to use the model GPT-4o Or GPT-4-Turbo,

“System Prompt” Strategy

We’re not going to just call it “writing blog posts.” The result is generic, robotic fluff. we need to use context injection,

We need to tell the AI Who This is and How Write.

Copy this into the system message field:

You are an expert content strategist and technical writer. You prefer short, punchy sentences. You should avoid corporate jargon like “unleash,” “leverage,” or “digital landscape.”

Your goal is to take the provided transcript and restructure it into a high-value academic article.

Keep the author’s original opinion and strong stance.
Use analogies to explain complex topics.
Use Markdown formatting (H1, H2, bolding).

Copy this into the user message field:

Here is the transcript: {{text_from_whisper_module}}

By explicitly restricting specific keywords and defining personality, the output changes from “general AI” to “professional draft”.

Step 4: JSON Secrets (Structure of the Data)

If you stop at step 3, ChatGPT will spit out a long block of text. Then you need to copy the title and paste it into Notion. Copy the body, paste it into Notion. Copy the LinkedIn caption, paste it into LinkedIn.

That is manual work. We hate working with our hands.

We are going to force the AI to give us data in a format that computers understand: JSON,

advanced signals

Go back to your OpenAI module (Step 3). We need to change the instructions. We don’t want just one article; We want a data object.

Add this to your user messages:

Important: You must output your response in a completely valid JSON format. Do not add any interactive text before or after the JSON. Use this schema:

{“blog_title”: “A catchy, SEO-optimized title”, “blog_body_markdown”: “Entire article content in Markdown”, “linkedin_post”: “A short, catchy LinkedIn post summarizing the article”, “tags”: (“tag1”, “tag2”, “tag3”) }

parser

Now, AI will return code, not text.

Add a JSON Module in make.
choose Parse JSON,
Map the output content from the OpenAI module to this parser.

Now, Make.com sees blog_title And linkedin_post As separate, draggable variables. This is the difference between “using AI” and “building software”.

Step 5: Landing (Notion Database)

Ultimately, we need to park this content where you can actually review it.

Add a perception Module: create a database item,
Select your content calendar database.
Map fields:

name Title): drag in blog_title Variable from JSON module.
page content: drag in blog_body_markdown Variable.
Situation: set it “draft”(Never set automation to “Publish” automatically, You are the editor; the AI is the writer),
LinkedIn Caption: map out linkedin_post Variable for text property in Notion.

outcome

Here’s the workflow in real life:

You end a client call where you’ve explained a complex topic perfectly.
You save the recording.
You drop the MP4 into your _INPUT_RAW_VIDEO folder.
You take lunch.
When you come back, you open Notion. There is a new page in your “Drafts” column. It has a headline, a full article, and social media snippets ready to use.

80/20 reality check

Let’s be honest about what we’ve created right now.

This system does not replace you. It replaces 80% of the earlier work. The output you get in Notion will be good, but it will not be perfect. This can misinterpret a joke. It may be using a metaphor that doesn’t quite fit.

Your work is no longer “authored.” Your job is “Editor in Chief”.

You have to read the draft. Add your personal flair. Get the rhythm right. But editing a 1,500-word draft takes 15 minutes. It takes 4 hours to write it from scratch. That is victory.

This workflow eliminates “blank page syndrome”. This ensures that your best ideas don’t die in your mind just because you didn’t have time to type them up.

next 15 minutes

Don’t try to build the entire spacecraft at once. start small.

Now, go to Google Drive and create it _INPUT_RAW_VIDEO folder. Then, log into make and just set up step 1 (trigger) and step 2 (transcription).

Even if you receive an automated email with a transcript of your voice notes, you’re already ahead of the 99% of people who rely on memory alone.

Automating Your Visual Presence

We have solved the problem of literal Material. We took your voice and turned it into blogs and LinkedIn posts. But in 2025, text is only half the battle.

The “video-first” philosophy implies that video is the last asset. But creating high-quality video for each use case – like a customized video resume for a specific job application, or a customized pitch deck video for each investor – is impossible to do manually. You can’t sit in front of a camera 40 hours a week.

Just as we used Make and OpenAI to automate the writing process, there are tools designed to automate the same picture Display.

If you’re looking to grow your personal brand beyond what can be physically recorded, you might want to explore portfoliovideoIt’s a platform that allows you to turn documents – resumes, pitch decks, scripts – into professional videos using AI avatars,

Imagine the workflow we just created, but in reverse: You feed a script into the system, and instead of a blog post, you get a studio-quality video of an avatar presenting your ideas, or a cinematic “Vibe Motion” header for your LinkedIn profile.

Automation is not about being lazy. It’s about making sure your ideas go further than your physical time allows.

Now, go have a rant. Your system is waiting.

Published via Towards AI