This is getting closer to the system I've been saying I wanted since the 1980s, whenever people said "what could you use a faster computer for anyway?"
It's a system where you can talk to it and make a photo realistic movie. The example I always use is, you're sitting at the computer, looking at a blank screen and you say something like:
"Ok opening scene. Dockside, London, early 19th century. Early evening. There are several ships docked, one being offloaded. Stevedores are working, some disreputable louts hanging around."
The screen is updating as I'm talking.
"OK make it grittier, more dirt and grime, let's have a fight break out in mid distance left of the screen. Now pan slowly right to reveal a bar called the Skull and Crown. Make the sign dirtier but let the last light of sunset glint off of the skull."
Screen updates. We are looking at what appears to be a Hollywood level period set full of extras who look the way they should, based on historical data that the model has.
"As we pan over towards the door Micky gets tossed out by the big burly barman. Make him younger, skinnier, he's about 17 years old."
The point of all this is, no, you don't need exact language to specify what you want. In the world of filmmaking you never do that. The screenwriter describes things in some detail, but always leaves a lot up to the interpretation of the director, the set designer, the costumer, the makeup artist, the casting director, etc.
The AI can take on any or all of those roles for us.
What I want is something I can control to make the movie I want to make. Then I want to be able to iterate on it: Let's make the main character a woman. Now everything gets changed to fit that. etc.
Of course the AI can replace the role of the writer too, and the director, and the producer leaving me with nothing to do. But the fact that I can bring my vision to the screen still makes it a great tool.