The easiest to use, because it lives inside ChatGPT and understands ordinary, messy prompts without the keyword incantations the others reward. You describe what you want in a sentence, then refine the result through conversation. It is at its best when you want one good image fast rather than a hundred variations.
It is not the sharpest on photorealism and gives less control than a dedicated tool, but the prompt understanding and the near-zero learning curve make it the natural pick for anyone who is not a hobbyist.