Image Generating Engine cannot follow instructions.

I’ve spent a couple unsuccessful hours trying to get the system to generate an image. The image is a split screen with a pane on the left and a pane on the right. The is a person in each frame. The person in the left frame must look left and the person in the right frame must look right. The system cannot get this consistently right. In fact it almost always gets it wrong. There seem to be an issue with how the instructions are passed or how/if they are interpreted by the image generating AI.