How to use image-to-image generation with DALL·E 3 via OpenAI API?

API
, , ,
1

Hi everyone,
I’m trying to use image-to-image generation with DALL·E 3 via the OpenAI API — similar to what’s possible inside ChatGPT (for example: “Anime style ”).

However, from what I can see in the official OpenAI API documentation, only supports text-to-image, and there’s no option to upload or reference an image as input for modification or context.

So here are my questions:

  1. Does the OpenAI API currently support image-to-image generation with DALL·E 3?
  2. If not, is this feature planned for release?
  3. Are there any recommended workarounds — for example, using gpt-4-vision to analyze the image and generate a prompt for DALL·E?

If anyone has working examples or official insights, I’d really appreciate the help. Thanks in advance!

2

Welcome, @geoor …

The style transfer is done with the 4o model which isn’t available in the API yet.

Hope that helps!

1 Like
3

Thanks ! In general, if we talk not only about the style, but about editing/improving the image/images in general, is it planned to be released ?

1 Like
4

Yeah, DALLE2 had inpainting which allowed you to edit a single part. DALLE3 didn’t have that feature, but it’s back in 4o…

DALLE was a model for just images while 4o is a “multi-modal” model meaning it’s been trained on text, audio, images, etc.

No word on when it will be available in the API, but you can take it for a spin on ChatGPT.

1 Like