Hyper-Scaling Ad Creative Generation using GPT and Fal.AI - n8n Workflow

• Marketing Agencies & Teams: Users needing to A/B test a large volume of visual ad creatives quickly.


• Performance Marketers: Professionals looking to scale campaign assets without manual creative work.


• n8n Automation Experts: Those seeking advanced examples of integrating complex AI models (OpenAI, Fal.AI) within an n8n workflow.


• Content Developers: Anyone interested in powerful n8n templates for high-throughput AI tasks.

Generating a high volume of visually distinct, yet strategically aligned, ad creatives is typically time-consuming and expensive. This powerful n8n workflow solves this by merging advanced AI capabilities into a seamless automation pipeline.

It starts with a single inspirational image and a user-defined creative 'recipe' (e.g., The Metaphor Bake, The Comparison Combo). The n8n automation first employs an OpenAI n8n node (GPT-4o) to deeply analyze the source image. Subsequently, a highly engineered GPT-5 system prompt is used to rewrite and generate hundreds of unique text prompts, ensuring compliance with the chosen creative recipe and user instructions.

These generated prompts are then fed into Fal.AI’s Nano Banana image editing model in batches. The workflow handles the asynchronous nature of image generation using a loop and polling mechanism before finally delivering the high-volume output files to a designated Google Drive folder. This setup transforms a manual, weeks-long design task into a fast, repeatable n8n template execution, significantly reducing time-to-market for new ad campaigns.

This comprehensive n8n workflow executes in five main stages:

1. Trigger & Input: The process begins with an n8n trigger, specifically the 'On form submission' n8n node. The user submits the company details, desired ad quantity, a reference image, and a creative recipe via a custom n8n form.


2. Image Analysis (GPT-4o): The uploaded image is processed. The 'Describe ad selected' n8n node (using GPT-4o) analyzes the image file and creates an exhaustive text description.


3. Prompt Generation (GPT-5): The core intelligence, the 'Create variants of the prompt' n8n node (using GPT-5), takes the image description, the user's creative recipe, and brand context to generate a JSON array containing the specified number of highly varied text prompts, ready for image generation.


4. Batch Generation & Polling (Fal.AI): The workflow splits these prompts into individual items using a Split Out n8n node, then enters a loop. For each variant, it calls the Fal.AI Nano Banana API via an HTTP Request node to submit the image editing job. Because generation is asynchronous, the flow uses a 'Wait 10 seconds' n8n node and then polls the status until the image is reported as COMPLETED.


5. Delivery: Once complete, the generated image URL is retrieved, the file is downloaded using an HTTP Request n8n node, and finally uploaded to the designated output folder in Google Drive. The loop then continues until all requested ads are generated.

To deploy this powerful n8n workflow, follow these setup steps:

1. Import: Copy the provided n8n workflow JSON and import it into your n8n instance.


2. n8n Form Trigger Setup: Open the 'On form submission' n8n trigger node and copy the URL. This is the endpoint your team will use to initiate the automation.


3. OpenAI Credentials: Update the credentials in both the 'Describe ad selected' and 'Create variants of the prompt' n8n node. You will need an API key connected to an account capable of accessing GPT-4o and GPT-5.


4. Fal.AI Setup: Sign up for Fal.AI and obtain your API key. In the 'Submit Request to generate image', 'Get Image status', and 'Get image url' n8n node, replace APIKEY> in the Authorization header with your actual Fal.AI key (prefixed with Key ).


5. Google Drive Setup: Update the Google Drive credentials. Specifically, ensure the folderId is correctly set for the input folder (in the 'Upload file' n8n node) and the output folder (in the 'Upload generated ad to output folder' n8n node). This n8n template requires read/write access to these folders.

On form submission (n8n trigger): This initial n8n node collects all required data: Company Name, number of ads, inspiration image (file upload), creative recipe dropdown, and specific user instructions. It acts as the gatekeeper for starting the n8n workflow.
Describe ad selected (OpenAI): Uses the GPT-4o model to perform precise visual analysis on the base64-encoded reference image, producing a detailed textual description that informs the subsequent prompt engineering step. This high-performance n8n node is crucial for accurate analysis.
Create variants of the prompt (OpenAI): Utilizes the GPT-5 model with a complex, structured system prompt. It takes the original image description and applies the selected creative methodology (e.g., metaphor or comparison) to generate a JSON array of image generation prompts. This is the core logic engine of the n8n workflow.
Split Out variants (n8n node): Splits the JSON array of generated prompts into individual items, allowing the n8n workflow to process each ad variant sequentially.
Submit Request to generate image (HTTP Request): Sends the modified prompt and the original image data (Base64) to the Fal.AI Nano Banana Edit API, initiating the image modification process.
Wait 10 seconds (n8n node): Essential for handling the asynchronous image generation. This n8n node ensures the workflow pauses before polling the Fal.AI status.

• Check if image is ready (If n8n node): Implements the polling loop by checking the job status. If the status is not COMPLETED, the n8n workflow loops back to the 'Wait 10 seconds' n8n node.