Aren't the fundamental pieces already there?

1 Generate a bunch of characters or objects based on a prompt.

2 Pick the one you like

3 Tell the AI to extract the character's traits from that one picture

4 Miracle happens (I don't know. What does a "character definition file" look like?)

5 Make a new prompt, but add the character definition from step 4, so you get the same traits, only in a different setting or position, etc