I tested the same concept again with MiniMax H3 Max, but this time in a completely different environment.
The interesting part is that the model still kept the hand-drawn effects, transformations, and camera tracking consistent with the new scene.
So this is basically the same prompt structure, just adapted to a different location.