Better AI Generate

Thank you for these suggestions.

Regarding point 1: This is now addressed in Firefly's Edit Image feature. The Markup tool allows you to identify and annotate specific objects in the image before generating, giving the AI the contextual anchoring you described and producing results that are more coherent with the existing scene.generative operation, giving the AI stronger spatial context and producing more scene-appropriate results.

Regarding point 2: The three-option selection model in Generative Fill has since been replaced with a single generated result. If the result is not satisfactory, you can discard it and regenerate without losing your original image. While this is a different approach to the workflow you described, the underlying concern — avoiding unwanted divergence between generations — is partially addressed by the ability to iterate from your original rather than choosing between multiple divergent options.

Please note it is best to include only one idea per post to track them through the ideation process seperately.

  droopy

There is already an similar option for this.
When talking the selection with the brush, reduce its opacity.
If the opacity is 100% the AI will generate an entirely new image.
If the opacity is 0% the AI will preserve the original image.
Using values like 40% to 70% you control how much the original image will be preserved and how much the AI will build on top of it.
The same goes for already generated images, just go to the generative fill and paint with the desired opacity to create variations of the same image.

Hi,

Yes, I follwed this way for submit the idea.

Thank you

Hi,

that sounds like an interesting proposal. Did you submit the idea via Firefly already?

Hope that helps.