Abstract/Summary

It is a premise that would have been met with derision several years ago, but the recent rise of diffusion models and their ability to produce hyperrealistic images now prompts us to engage seriously with this question. Here, we use three diffusion models – DALLE-2, Midjourney, and Stable Diffusion – to explore whether neural networks can produce realistic photographs of weather phenomena, and how to use their strengths in a manner that might win a weather photography competition. Finally, we invite the reader to take part in a test and judge which photos were previous finalists, and which were AI-generated.