Generative Fill prompts to recognize text-method alignments better

Extend Generative Fill (Firefly) to understand natural language to text-method alignments in addition to text-image alignments. I.e. in addition to "what" kind of imagery to generate, "how" it should be accomplished.

I'm sure that already happens to some extent in just "what", so in pics:

"very smooth transition/morph/blend/mix between <left and right textures described in detail>" or whatever else you try:

= Not good.

Proposal: Generative Fill would have some understanding of what I'm trying to achieve when I'm specifying methods to be applied to already-existing imagery, rather than new content/imagery. I hope this is understandable.

It's a generic problem in state-of-the-art AI-image generators, but in my mind addressing it would be particularly beneficial for Adobe.