Is Long-Term Consistency achievable?

[bondobrus]

Hi! I've been playing with svi for few weeks and I really love it. I found one issue I'm struggle with since this time - it's context consistency in multiple generation parts. I'm using this workflow

Example: The camera is fixed and my character is dancing there's door in background (not vissible in reference image) and the doors are not vissible at the end of first part again. On the 2nd part of the video character is still dancing but in same place where doors should be there's wall.

Question: Can I somehow provide "whole context of first part" for next part of generation? Thank you guys