The_Chosen_One:_Consistent_Characters_in_Text-to-Image_Diffusion_Models