This is impressive. I'm working in an adjacent space and one thing that has been a challenge is character consistency -- having characters in AI-generated art a

This is impressive. I'm working in an adjacent space and one thing that has been a challenge is character consistency -- having characters in AI-generated art appear reasonably similar across different image generations. Your implementation looks to do a solid job -- any learnings that you're willing to share?