🎬 Create Instagram carousel posts using GPT-4.1 Nano and Gemini 2.5 Flash

⚡ 666 views · 🎬 Content Creation & Video

💡 Pro Tip — Heads up — if this workflow’s Instagram scraping breaks (it happens whenever Meta updates their frontend), ScraperNode is a community node that handles profiles, posts, reels, and comments without any HTML parsing.

View All Scrapers

Description

Tired of manually creating Instagram carousel content? This workflow takes a single content idea from a database, uses GPT-4.1 Nano + Gemini 2.5 Flash to generate unique images and captions, uploads them, and publishes a fully formatted Instagram carousel — all without touching your phone.

🧠 How It Works

Feed it a topic. It does the rest.

🔁 Workflow Breakdown

1. ▶️ Manual Trigger — When clicking 'Execute workflow'

Kicks off the workflow on demand. Swap this with a Schedule Trigger or Webhook to make it fully hands-free and time-based.

2. 📋 Get Row(s) — Data Table (Get)

Fetches the next unprocessed content idea from an n8n Data Table. It filters rows where the status column is empty, ensuring only fresh, unpublished ideas are picked up. Each row contains:

3. 🧩 OpenAI Chat Model — GPT-4.1 Nano

Acts as the language model backbone for the prompt engineering agent. Connected to the Generate-Prompt agent as its AI brain, it powers all text reasoning and JSON output generation.

4. ✍️ Generate-Prompt — LangChain AI Agent

An expert prompt engineering agent powered by GPT-4.1 Nano. It takes the core idea and produces a structured JSON array containing:

It outputs N distinct prompts based on the number_of_images field.

5. 🔄 Parse_Array — Code Node

Parses the agent’s raw JSON string output into proper n8n items, one per image prompt. This allows the loop below to process each prompt individually. Without this step, the array would remain a single unprocessed string.

6. 🔁 Loop For Each Prompt — Split In Batches

Iterates over each image prompt one at a time. On the final iteration, it routes to the Collect Image URLs node to assemble everything. On every other pass, it routes to Generate an image to keep generating.

7. 🎨 Generate an Image — Google Gemini 2.5 Flash Image

Uses Gemini 2.5 Flash Image (Nano Banana) to generate a high-quality image for each prompt. The detailed prompt from the agent is passed directly here, ensuring each carousel slide is visually distinct from the others.

8. ☁️ Upload a File — Upload To URL

Takes the generated binary image and uploads it to a publicly accessible URL via the uploadtourl service. Instagram requires all carousel images to be hosted at public URLs before publishing — this node handles that requirement.

🔑 Get your free API key here: uploadtourl.com

9. ⏱️ Rate Control — Code Node

Two critical jobs in one node:

  1. Accumulates the uploaded image URL into global workflow static data so URLs persist across loop iterations.
  2. Waits 15 seconds between iterations to respect Instagram and image generation API rate limits, preventing failures during bulk runs.

10. 📦 Collect Image URLs — Code Node

After the loop completes, this node retrieves all accumulated image URLs from global static data, resets the store for the next run, and bundles everything into a single object:

This is the final assembly step before publishing.

11. 📸 Publish on Instagram — Instagram Node

Publishes the full carousel post to Instagram using the Facebook Graph API. It sends all image URLs in one request along with the AI-generated caption. Targets a specific Instagram Business Account by node ID.

12. ✅ Update Row(s) — Data Table (Update)

After successful publishing, marks the row in the Data Table as Published, writing in the returned post ID. This prevents the same idea from being posted twice and keeps your content queue clean and trackable.

🛠️ Tech Stack

⚙️ Setup Requirements

🔗 Nodes Used

AI Agent, OpenAI Chat Model, Google Gemini, Data table

📥 Import

Download workflow.json and import into n8n: Workflow menu → Import from File

📖 Importing guide · 🔑 Credential setup