Thought it would be a machine learning text to image based on what is being described. Now my brain is stuck on what a speech to text to image model would generate when being driven by a dungeon master.