Your text prompts should guide the model towards what part of the image you want to focus on. Basically the message beneath involves guiding the model via text to artifically increase the resolution of the image.