Recently, I've been seeing posts from people that show Stable Diffusion images that look a lot better than what Midjourney spits out (particularly if they involve text), so I thought I'd give it another go.
On the left, an image from September, 2022, produced by Stable Diffusion when given the prompt "Book cover. Man speaking to a blonde woman in Oxford". As you can see, it has some ... problems. On the right, the result of the same prompt from this morning.
Well they have fewer arms and shorter necks, but the eyelines need work.
For comparison, here's what Midjourney gave me:
The eyelines aren't a lot better, and it seems to favour a particular kind of book. The uncanny valley doesn't kick in, though.
Hmm. For the time being, I think I'll stick with Midjourney for doodling with during interminable Zoom meetings.