There have been a couple attempts at this kind of thing (same with AI generation of images from pages of Choose-Your-Own-Adventure books).

It's more a gimmick than anything particularly useful. Might even distract if the image embellishes from the original description leading players down the wrong path for solving a puzzle.