When you submit an instruction to Art Director, we use an LLM to modify the original prompt for the selected image, and then create a new image based on the updated prompt. For example, if you had selected an image with the prompt:
A person is floating against a backdrop of a bright blue sky.
And you then provided this instruction to Art Director:
Make it nighttime
The revised prompt might be:
A person is floating against a backdrop of a nighttime sky.
Because you are not editing the image directly, but instead editing the prompt that was used to generate that image, the new image may change in more ways than intended. We are aware of this limitation and are working on ways to make changes more precise, so only the specific elements you request will be altered.