Lazy AI - not as good as Microsoft Copilot at following instructions

I provided the following prompt:

A photorealistic view of the inside wall of a tornado from the vantage point of the opposite inside edge of the ring of the tornado - looking across the horizon - not up into the sky. No objects flying around in it. I just want the swirling cloud. Make it lightly shaded and mostly shades of gray.

It just produced a generic picture of a tornado in the distance.  One tornado, it produced sideways.  So, I followed up with this prompt:

A photorealistic view of the inside of a tornado. As if I were standing inside the tornado looking at the wall of cloud sweeping across my view. I want it to be clean with no objects in it.

It literally posted the same tornados as the first time with little variation.  

Imagine I'm standing inside a tornado - during the quiet part.  Instead of looking up at the sky into the top of the funnel, I'm looking across to the opposite wall of wind/clouds swirling around me.  I just want to see the swirling inside wall of clouds - not a generic tornado in the distance.  What else can I sat to prompt it so that I'm not using up credits on a bot that can't follow instructions beyond the word "tornado"... .