Also now for the prompt "Mature oranges on the table", giving an image of example in "Style" section and selecting image as photo, it generate an image that isn't as a photo and nothing have considered the example image added in "Style" section.
Adobe Stock is a significant source of training images for Firefly. If you search for Clementines in Adobe stock, you can see many of those images have leaves, so I think that is where the model is getting this from:
I was not able to completely eliminate the leaves. The prompt that generated the least leaves was "picked Clementines in a bowl, simple". I think your best (fastest) solution is to generate with the least leaves and then take to generative fill to remove the leaves in the image.
Thank you again for the detailed feedback and examples.
We've shared your case with our engineering team for further investigation. As soon as we receive an update from them, we’ll be sure to let you know right away.
We appreciate your patience and are here to support you.