Competitor image ad rebuilt around your product

Competitor image ad rebuilt around your product

An n8n workflow with a three-field form. You send the picture address of a rival's ad, a photo of your product and any
wording you want changed. Gemini 2.5 Pro reads both pictures and plans the swap, and a Gemini image model draws the same
ad with your packaging in it. The picture stays in the run as a file.

What is inside

• workflow.json: the workflow as n8n exports it, eleven nodes, one of them a sticky note that holds only the workflow's
  title. It holds no credential.
• README.md: these notes.

What we changed from the original copy

• generate_ad_image called gemini-2.5-flash-image-preview, which Google shut down on 15 January 2026. It now calls
  gemini-nano-banana-2.1, the replacement on Google's deprecations page, read on 3 November 2026.
• Both Gemini requests were linked to a credential in their author's n8n. The link is gone, so select your own in each.
• Nothing else. The file has no Code node, and its expressions needed no repair.

Before you start

• An n8n instance, on n8n Cloud or your own server.
• A Google AI Studio key for the Gemini API, with access to gemini-2.5-pro and gemini-nano-banana-2.1.
• The address of the rival's ad picture itself, not of the post around it. A JPEG matches the request as it stands.
• A photo of your product, best as a PNG with the packaging easy to see.

Set it up

1. In n8n, create a workflow, open its menu and choose Import from File. Pick workflow.json.
2. Add a Header Auth credential. Set its name to x-goog-api-key and its value to your Gemini key. Select it in
   generate_ad_image_prompt and generate_ad_image.
3. Open build_prompt and fill the gaps listed under "The prompt's gaps" below.
4. If you will share the form, rename it in form_trigger. Its title is Facebook Ad Thief Form.
5. Save, open the form's test address and send one ad and one photo.
6. Press Publish, then share the form's production address.

The form confirms at once, and the run goes on behind it. Open the run under Executions and download the picture from
get_image.

The prompt's gaps

build_prompt holds the instructions for Gemini 2.5 Pro. The brand names were cut out of its copy, so a few
sentences stop short:

| Where in the text | What goes there |
|---|---|
| After text that says and labels on products that say | Your rival's brand name, as it shows on their ads |
| After promotes our product | Your product's name |
| After switching competitor brand name to our name | Your brand name |
| The fragments "A", "AG", "G" and "G1" | Pieces of your rival's name that a cropped label might show |

The prompt runs without them. Gemini then has to guess which name to remove and which to write.

How a run goes

1. form_trigger (Form Trigger 2.2) takes three fields: Competitor Image Link, Your Product Image (one file) and
   Additional changes (optional).
2. convert_product_image_to_base64 turns your photo into base64 text. download_image fetches the rival's picture
   from its address, and convert_ad_image_to_base64 turns it into text too.
3. build_prompt sets the instructions, with your Additional changes at the end.
4. generate_ad_image_prompt sends the instructions and both pictures to gemini-2.5-pro. Its answer is the edit plan.
5. generate_ad_image sends that plan and both pictures to gemini-nano-banana-2.1, which answers with a picture.
6. Wait pauses 5 seconds. set_result keeps the picture's base64 data, and get_image turns it into a file named
   data.

Change the look

• The plan: the instructions in build_prompt. As written, they ask for a near copy of the rival's ad, with the brand,
  the packaging and the words you name changed.
• Models: the addresses in generate_ad_image_prompt and generate_ad_image.
• Picture types: mime_type in the JSON body of both requests, image/png for your photo and image/jpeg for the ad.
• The form: the title, the description and the field labels in form_trigger. If you rename Your Product Image, set
  binaryPropertyName in convert_product_image_to_base64 to the new label, with underscores for spaces.
• Where the picture goes: add a node after get_image, such as Google Drive, Gmail or Slack.

Good to know

• The link must open the picture itself. A Facebook post's address returns a web page, which Gemini cannot read as an
  image. Open the ad's picture on its own and copy that address.
• Both requests call your photo a PNG and the ad a JPEG, whatever you send. Send those types, or change mime_type.
• Wait comes after a call that has already answered, so it only adds 5 seconds. You can delete it.
• Google's deprecations page gives 1 December 2026 for gemini-2.5-pro. From that day Google plans to send its requests
  to gemini-3.8-flash, so change the address in generate_ad_image_prompt first if you want to pick the model yourself.
• Each run is two Gemini calls: one reads two pictures and writes a plan, the other draws a picture.
• Once the workflow is published, anyone with the form's address can start a run on your key.
• The rival's photo, layout and words belong to them. Change enough to make the ad your own, and check the advertising
  rules where you sell before it runs.
• Both pictures and your notes go to Google.