Give the generative ai a memory to give the user the ability to get various versions of a set scene then using future prompts to be able to alter the selected scene/format
@Alan37278164uujb what do you mean by "memory"? Have you tried using Structure Reference as a guide? Generated your set scene, then use it as a structure reference to alter further with prompts...
Similarly to using a regular chatgtp bot the bot remembers your previous prompt and its own generative text. In this case for image generated ai you'd be able to select the exact picture you want with all its components then make minor adjustments using other prompts without completely changing the overall base picture. It may be more complex for the ai at this point to achieve keeping all the main aspects of the original/generated picture that is desired but overall I feel this is what would make the difference in the utility of generative ai for users