Try something like 'a photo of a man/woman/person'
What the model should NOT produce.
Width of the generated image
Height of the generated image
How much the written prompt weighs into the generated images.
How much your uploaded files weigh into the generated images.