Using gpt-image-1 in n8n to generate composite game cover from multiple character images?

[Brett_Hoy]

Hi all,

I’m currently exploring how to work with image inputs in gpt-image-1 using n8n, and had a use case I wanted to validate:

Goal:

I’d like to provide multiple character images (e.g., 3–5 separate portraits of characters) and have the AI generate a single composite image — something like a game cover or poster that includes all of them, properly arranged/stylized.

What I Know:

• This can be done using the OpenAI API in Python by uploading multiple image references and prompting GPT to create a merged visual output.
• I can work with OpenAI nodes or the HTTP Request node in n8n to hit their API endpoints.

My Questions:

1. Is it currently possible to do multi-image input → single image output using OpenAI’s image generation tools directly within n8n?
2. If yes, what’s the best workflow for handling multiple image inputs? Would this involve storing them in a binary format, referencing image URLs, or something else?
3. Has anyone done something similar using gpt-image-1, or similar for multi-character composition workflows in n8n?

Any guidance, templates, or best practices would be hugely appreciated!

Thanks in advance

[tridi]

Hi! Coincidentally, I’m working on something similar. I rely on image input from S3 storage or Dropbox because, via the n8n HTTP Request node, the only input allowed by OpenAI is binary.

If you need more than one image reference, you’ll need to combine them with a Merge node before inputting them into the HTTP Request. Below is the HTTP Request node I use:

[Brett_Hoy]

Hi Tridi,

Really cool to hear you’re working on something similar!

Here’s what I’ve got so far:

I’ve set up a loop that iterates through a dataset of prompts — each iteration generates an image. I think the images are already merged by the loop? Not too sure exactly how to pass those images on to GPT-Image-1. Would you mind giving some more clarity please?