I wonder if anyone here can offer some advice. . .

When I create an AI video involving mess, the mess seems to disappear from the victim too fast.

Example A: A pie-in-the-face aftermath. The face has just been pied but now appears clean. I have worded the prompt as best as I can to keep it messy, but it tends to default to clean.

Example B: A nude woman falls on her back in mud and is nicely covered. But if I have her roll onto her stomach, her back is suddenly perfectly clean and dry. Again, I prompt for a coating of mud to remain on her back but AI seems to have deaf ears in this area.

Has anyone else experienced this and perhaps can offer some advice?

My example photo shows what I mean, the woman was face down in the mess, but when she rolls over, the front of her body is mysteriously clean which I do not want.

Not sure if this will help but with the way Ai works by being prompted my words maybe saying 'the mess does not dissipate afterwards' or something like that might work.