Reply to post:

If I can generate the same image by a purely descriptive text prompt of 10 or so words, that isn't itself copyrightable, then the image cannot be copyrightable. Why should the first person to feed those words into a model be entitled to a copyright?