In a lot of my projects I ask ChatGPT 4 to create images.
If I like the image but want a minor change, for example, to remove the text or to make the character change direction or to increase the width of the image, ChatGPT will then generate a completely new image.
The image edit technology of “gpt-4o image-1 imagen whatever” they’ve got in ChatGPT cannot edit images.
ChatGPT used to have DALL-E 3 that could edit its own AI images using a drawn mask, but that is now gone also and was never released on the API.
You observe correctly, OpenAI is stretching the lack of truth whenever they suggest that this image model can preserve the original in any way other than by visual impression or even produce something that is useful to blend in manually in your own photo editing software.
This topic was automatically closed after 22 hours. New replies are no longer allowed.