For bing the prompt is very important because you need the right verbiage to fool the filters, but it's also a bit random from what I've seen in the sense that some accounts trip the dog and others don't.
In stable diffusion the prompt isn't as important as choosing the right checkpoint, loras, and other settings. Also you need to try a shit ton of times to get what you want or something similar at least. The prompt can say a lot of things but the models aren't as powerful as what dall-e might have in the background, since this is something that needs to run in a limited amount of memory.
See the image below, the prompt is "a handsome man swimming in a pool"
It actually doesn't help you in any way because you need all the extra model details and to find the right seed to get that exact image but sometimes even with that you might not get the same one.
In stable diffusion the prompt isn't as important as choosing the right checkpoint, loras, and other settings. Also you need to try a shit ton of times to get what you want or something similar at least. The prompt can say a lot of things but the models aren't as powerful as what dall-e might have in the background, since this is something that needs to run in a limited amount of memory.
See the image below, the prompt is "a handsome man swimming in a pool"
It actually doesn't help you in any way because you need all the extra model details and to find the right seed to get that exact image but sometimes even with that you might not get the same one.