Nano Banana Influencer Ad Creative Generator

Use n8n, Google Drive, and Google Gemini to automatically generate influencer-style ad creatives at scale. This instructional guide walks you through what the Nano Banana workflow does, how each node works, how to set it up, and how to write strong prompts for consistent influencer ad images.

What you will learn

By the end of this guide, you will be able to:

• Explain why automating influencer-style ad creative is useful for marketing and testing.
• Understand the core components of the Nano Banana n8n workflow template.
• Configure Google Drive and Google Gemini for image generation in n8n.
• Follow the workflow node-by-node to see how product and influencer images are combined.
• Write effective prompts for influencer-style ad images.
• Troubleshoot common issues and optimize the workflow for performance.

Why automate influencer ad creative?

Influencer-style content tends to convert well because it looks personal, relatable, and less like a traditional ad. The challenge is scale. Creating many variations of influencer photos with your product usually requires:

• Coordinating with multiple creators.
• Organizing photo shoots.
• Editing and resizing images for each channel.

This is time-consuming and expensive, especially if you want to A/B test different angles, backgrounds, and styles.

With an n8n workflow and an image generation model like Google Gemini, you can:

• Automatically combine a single product image with many influencer reference images.
• Generate consistent, on-brand ad creatives faster.
• Maintain control over style, pose, and overall composition through prompts.
• Support use cases such as social media posts, paid ads, and creative testing.

Core components of the Nano Banana workflow

The Nano Banana Influencer Ad Creative Generator is built around a simple idea: take one product image, combine it with multiple influencer reference images, and generate new influencer-style ads automatically.

Key tools and technologies

• Platform: n8n workflow automation
• Storage: Google Drive (one folder for source influencer images, one for generated images)
• Model: Google Gemini image generation endpoint
• Image handling: base64 encoding and binary conversions
• Primary use cases: influencer ad creative, social content, A/B creative testing

How the workflow works in n8n

At a high level, the workflow:

1. Receives a product image (for example via a form).
2. Converts the product image to base64.
3. Fetches a list of influencer reference images from Google Drive.
4. Loops through each influencer image, converts it to base64, and sends both images to Google Gemini.
5. Receives the generated influencer-style ad image, converts it back to a file, and uploads it to a destination Drive folder.

Below is a node-by-node explanation so you can follow the data flow clearly.

1. form_trigger – capture the product image

The workflow typically starts when a user uploads a product image through a form. In n8n, this is handled by a node such as form_trigger (or any other trigger node you prefer).

This node:

• Accepts a single file upload (your product image).
• Stores the file as binary data in the workflow.
• Makes this product image available to all following nodes.

The uploaded product image is the central element that will appear in every generated influencer-style image.

2. product_image_to_base64 – convert product image

Most image generation APIs expect image data in base64 format. The product_image_to_base64 node converts the binary product image from the trigger into a base64 string.

Key points:

• Ensure the binary property name in this node matches the file key from form_trigger.
• The resulting base64 string will be passed to the Gemini request as inline image data.

3. list_influencer_images – get reference images from Google Drive

Next, the workflow needs influencer reference images. These are photos of people who will appear to be holding or interacting with your product in the final generated images.

The list_influencer_images node:

• Connects to Google Drive.
• Lists all files in a specific source folder that you configure.
• Returns a collection of influencer image file IDs and metadata.

Set the folderId in this node to the Google Drive folder where you store your influencer reference images.

4. iterate_influencer_images and download_influencer_image – loop and convert

Now the workflow needs to process each influencer image one by one. This is typically done with:

• A loop or splitInBatches configuration (often referred to as iterate_influencer_images).
• A download_influencer_image node that fetches each image from Drive.

Within this loop:

1. The workflow takes one influencer image ID from list_influencer_images.
2. download_influencer_image downloads that file as binary.
3. A conversion step (for example an influencer_image_to_base_64-style node) turns the binary into base64, just like the product image.

Batching is important. By using splitInBatches or similar logic, you can:

• Control how many images are processed at once.
• Avoid overwhelming the Gemini API.
• Stay within Google Drive API rate limits.

5. generate_image – call Google Gemini

Once you have:

• The product image in base64.
• The current influencer image in base64.

you are ready to call the Google Gemini image generation endpoint.

The generate_image node is typically an HTTP Request node configured to:

• Send a POST request to the Gemini image generation endpoint.
• Include authentication with an API key or OAuth, depending on your Gemini setup.
• Pass both base64 images as inline data in the request body.
• Include a well-structured prompt that describes how the product and influencer should be combined (pose, background, style, etc.).

The node returns a response that includes the generated image in base64 format. Usually, you generate one final image per influencer reference.

6. set_result, get_image, and upload_image – save the generated output

After Gemini returns the generated image, you need to convert and store it.

• set_result: Extract the base64 image from the Gemini response and map it into a property such as image_result.
• get_image (convertToFile): Convert the base64 string back into binary file data so it can be uploaded to Google Drive.
• upload_image: Upload the binary file to your destination Google Drive folder.

In upload_image, set a consistent naming convention to keep outputs organized. For example:

Influencer Image #{{ $runIndex + 1 }}

This makes it easier to review, compare, and use the generated creatives in your campaigns.

Step-by-step setup guide

Use this section as a checklist to get the Nano Banana workflow running in your own n8n instance.

Step 1 – Prepare Google Drive folders

• Create a source folder in Google Drive for your influencer reference images.
• Create a destination folder where all generated influencer-style images will be stored.
• Ensure your service account or OAuth user has the correct permissions for both folders (read for source, write for destination).

Step 2 – Configure n8n nodes

In your n8n workflow, configure the main nodes as follows:

• form_trigger:
  • Set it up to accept a single file upload for the product image.
  • Note the binary property name used for this file.
• product_image_to_base64:
  • Use the same binary property name as in form_trigger.
  • Output a base64 string property for the product image.
• influencer_image_to_base_64 (or equivalent):
  • After downloading each influencer image, convert its binary data to base64.
  • Make sure the binary property name matches the output of download_influencer_image.
• list_influencer_images:
  • Set folderId to the ID of your source influencer images folder.
• download_influencer_image:
  • Use the file id from list_influencer_images as the input.
  • Output the file as binary.

Step 3 – Add Gemini / API credentials

To call Google Gemini from n8n:

• Use an HTTP Request node for generate_image.
• Configure the request URL to point to the appropriate Google Gemini image generation endpoint.
• Add authentication headers, such as:
  • An API key header, or
  • OAuth credentials, depending on your Gemini account setup.
• Include the product and influencer base64 images and your prompt in the request body.

Before scaling, test with a single product and influencer image pair to:

• Verify that authentication works.
• Check that the response format matches what your next nodes expect.
• Refine your prompt until the generated image looks correct.

Step 4 – Set file conversion and upload logic

Finally, configure the nodes that handle the Gemini response and upload the final images.

• set_result:
  • Extract the base64 image string from the Gemini response JSON.
  • Store it in a property, for example image_result.
• get_image (convertToFile):
  • Convert the image_result base64 string to a binary file.
  • Set the correct file type (for example image/png or image/jpeg).
• upload_image:
  • Point this node to your destination Google Drive folder.
  • Use a clear naming pattern, such as:

    Influencer Image #{{ $runIndex + 1 }}

  • Optionally add metadata for campaign, variant, or test group.

Prompt tips for better influencer-style ad images

The quality of your prompt has a major impact on the output. You are not just describing the scene, you are guiding the model on how to combine two specific images.

What to include in your prompt

• Pose and interaction:
  • Describe exactly how the person interacts with the product.
  • Example: “The person holds the tumbler to their lips as if about to take a sip.”
• Style and mood:
  • Set the emotional tone and visual style.
  • Example: “Natural, candid, warm lighting, smiling, friend-taken photo.”
• Camera angle and framing:
  • Specify where the camera is and what is visible.
  • Example: “Slightly angled from the side, 3/4 view, visible table and cafe background.”
• Clarity and brevity:
  • Keep prompts concise but descriptive.
  • Use clear constraints and avoid conflicting instructions.
  • Finish with a direct instruction like “Only return the final generated image.”
• Source image quality:
  • Use product and influencer images with decent resolution.
  • Sharp, well-lit input images help the model create cleaner composites.

Example prompt snippet

"Create an image where the cup from image 1 is held by the person in image 2. The person sits at a cafe table, smiling warmly at the camera. Natural, friend-taken photo style, slight side angle. Only return the final generated image."

Legal and ethical considerations

When you generate influencer-style images using real people as references, you are still dealing with real identities and potential brand impact. Keep these points in mind:

• Consent: Only use reference photos of people who have explicitly agreed to this type of use.
• Disclosure: Follow local advertising rules about sponsored content and influencer disclosures in the markets where your ads will run.
• Intellectual property: Make sure you have the rights to use both the product images and the influencer reference images, and check that your usage complies with the model provider’s policies.
• Privacy and reputation: Avoid generating images that misrepresent, defame, or otherwise harm the person in the reference photo.

Troubleshooting and optimization

Common issues and fixes

• Low-quality or unrealistic outputs:
  • Increase the resolution of your source product and influencer images.
  • Refine your prompt with more detail about lighting, angle, and style.
• API errors or rate limits:
  • Implement retry logic in your HTTP Request node.
  • Use splitInBatches to process images in smaller groups.
  • Respect both Gemini and Google Drive provider quotas.
• Mismatched background or composition:
  • Be explicit about the setting, for example “indoor cafe, shallow depth of field, blurred background.”
  • Specify how prominent the product should be in the frame.

Performance tips for scaling

• Batch processing:
  • Use splitInBatches or similar logic to throttle calls to Gemini.
  • Adjust batch size based on your API limits and desired throughput.
• Naming and metadata:
  • Add campaign names, variants, or test IDs to file names or metadata.
  • This makes A/B testing and performance analysis much easier later.