Auto-generate custom cake images with n8n, OpenAI GPT & Replicate Flux Schnell

User submits a name via Webhook → AI Agent generates a detailed prompt → Replicate renders the cake image → Image URL returned to frontend

Who

Bakeries, event planners, e-commerce platforms, or operators who need personalized celebratory visuals at scale

You get

Users enter a name and receive a custom AI-generated cake image featuring that name, returned in real time via Webhook

You need

n8n

Steps

1. Import the template

   Import template ID 6161 into n8n, or import directly from workflow.json. After import you should see 6 functional nodes and 5 sticky notes.

2. Configure API credentials

   Add two credentials in n8n Credentials: 1) OpenAI API Key, type openAiApi, for the OpenAI Chat Model node; 2) Replicate API Token, type httpHeaderAuth (Header Name: Authorization, Value: Token <your-token>), for the Generate Image node. Do not hardcode keys in node parameters.

3. Configure the Webhook trigger

   Open the User Sends Request (Webhook) node and confirm HTTP Method is POST and Response Mode is responseNode. Copy the generated Webhook URL and configure it in your frontend form or Bolt frontend, ensuring the form POSTs JSON containing body.name (e.g., {"body":{"name":"Sarah's Birthday"}}).

4. Configure the Prompt Generator node prompt

   Open the Prompt Generator (AI Agent) node, confirm promptType is define, and paste the following text into the text field. This prompt instructs the model to generate a detailed cake-image prompt based on the user's name, outputting only plain text with no quotes or extra information.

   Execution prompt · Source text。Prompt Generator node text field (promptType=define)

   Copy prompt

   =You are an expert in creating ai cake images.
   I want you to randomise the cake And put a very specific  {{ $json.body.name }} On the cake.
   
   But remember one thing I don't want you to add anything else other than the prompt In the output. I dont want you to add any quotes Or any other information other than the plain text of the prompt.
   
   Also wherever you mention to write the name
   Instead of "{{ $json.body.name }}" write it \"{{ $json.body.name }}\""
   the name \nabin\ is {{ $json.body.name }}
   
   eg : black forest gateau cake spelling out the words \"FLUX SCHNELL\"", tasty, food photography, dynamic shot
   
   give a very detailed prompt so that it can perfectly create the images of the cake

5. Configure the Replicate image generation node

   Open the Generate Image (HTTP Request) node and confirm: Method is POST, URL is https://api.replicate.com/v1/models/black-forest-labs/flux-schnell/predictions, Authentication is genericCredentialType / httpHeaderAuth, Body type is JSON, and body content is {"version":"latest","input":{"prompt":"{{$json['prompt']}}"}}. This node sends the previously generated prompt to Replicate Flux Schnell.

6. Configure the image download node

   Open the Download Image (HTTP Request) node and confirm URL is {{ $json.output[0] }} and Method is GET. This node extracts the image URL from the Replicate response and downloads it.

7. Configure the Webhook response node

   Open the Respond to Requests (Respond to Webhook) node and keep default parameters. This node returns the Download Image output as the response to the frontend. After configuration, activate the workflow and send a test request from your frontend form to verify the cake image URL is returned.

You are setting up an n8n workflow: users submit a name via Webhook, an AI Agent generates a cake-image prompt, Replicate Flux Schnell renders the image, and the URL is returned. First confirm you have an OpenAI API Key and Replicate API Token ready, and that you have imported template 6161. Then configure credentials, Webhook, Prompt Generator prompt, Replicate HTTP node, and test the frontend call.