The docs should clarify that inpainting with gpt-image-1 does not do a pixel replacement at the mask level like dalle-2 does but more of a soft mask with total image recreation.
Even looking more carefully now at the example of your press release: https://platform.openai.com/docs/guides/image-generation#edit-an-image-using-a-mask-inpainting
It’s obvious now that the entire image was re-generated.
My feedback would be to support the actual claimed technique of in-painting (hoping that you do 
) or please update the docs so future users are aware of this without feeling like there’s a bug or something and waste time trying to fix their masks like I did yesterday.
Thanks!
12 Likes
I am also seeing the same issue. The mask is not working at all.
The whole picture get’s regenerated, even though looking similar, but it’s not 100% same. Inpainting mask became useless.
They mentioned in another thread that they have it slated as a product feature but have no ETA for it.
Please follow here for a new update:
Gpt-image-1 problems with mask edits - #41 by OpenAI_Support
OpenAI_Support:
We have plans to implement precise in-painting, however, cannot commit to any timelines right now. Please stay tuned for updates!.
Gpt-image-1 problems with mask edits - #37 by OpenAI_Support
OpenAI_Support:
We’d like to provide an update to everyone on this thread: the behavior observed with gpt-image-1 has unfortunately been confirmed as a known limitation. While the functionality performs as expected when using dall-e-2 - albeit with some compromise in image quality - the issue described here is indeed present with gpt-image-1 . We sincerely apologize for the inconvenience this may have caused.