I agree. People aren't grasping the magnitude because they're thinking about jobs. Jobs a silly way to measure this. Jobs are temporary. Nobody worries about the mechanical stocking frame making socks anymore.
This is more like the literacy/printing press transition.
Used to be, people had to learn to memorize a lifetime of stories and lore. Now nobody learns to make a memory palace or form a mnemonic couplet. Why would you bother? You just write things down.
Today, people learn to draw. In a generation, why would you bother?
There will still be specialist jobs for people generating images, but instead of learning to make them up, the specialists will be very good at picking them, suggesting them, consuming them.
Humans will be the managers and the editors, not the creators.
The same thing will happen to other arts. First (and very easily) to music. Eventually, perhaps, to writing and whole movies...
The only thing stopping that is that the models can't maintain a reality between frames. They can't make an arc. It's all dreamlike.
If we find a way to nail object persistence it will be a singularity-level event. The moment you can say "make another version of this movie, but I want Edgar to be more sarcastic and Lisa should break up with him in the second act" we will close the feedback loop.
It's a lot bigger than "lost jobs".