So much human art will be enabled by this tech.
Folks bemoaning the death of creativity do not get it. This is a tool for turning shooting-scripts into finished shots. Idiot suits imagine they'll have a movie business without writers, when what this text-driven technology actually fucking allows is a movie business exclusively for writers. Alright, and editors. The model should be trained on more raw footage to avoid inserting its own jump-cuts.
Denoising is the heart of this technology. (Assuming it's like Stable Diffusion. It's hard to keep track.) That means human drawings, real actors, and manual MS Paint edits can be placed anywhere in the process, and the machine will plow ahead like that's its own output. This character's moustache is missing? Scribble scribble, there you go, looks flawless in the end.
What show do you want to exist? What movie's non-existence vaguely offends you? You can't type "Firefly season 2 good version 1080p" and expect that to just happen - but your fanfiction can now look like a DVD rip. You can have those characters look just right, without requiring Joss Whedon's rolodex, a mountain of cash, and a time machine.
And anything truly new, any weird story you want to tell, doesn't have to star actors. Your main guy doesn't need to look like any real person. You don't need a crew, or a schedule, or... catering. It's just you. It's all up to you.
And a machine that probably costs a thousand dollars an hour to run, but give it another year and any laptop will get better results.