This is the first time I generated an image in Midjourney where I actually couldn’t tell it was generated by a text-to-image model.

This is the first time I generated an image in Midjourney where I actually couldn’t tell it was generated by a text-to-image model.