What do I think about text-to-video LLM generation, represented by OpenAI's Sora?
I think you should read this:
And this:
When we were making our debut sci-fi short in 2014, I'd be thrilled to have something like that (here you can watch it and also learn about the whole process). By the moment it hit festival screens, I'd have been confident I wouldn't need such tool. Because you can put any movie star into your text-to-video, but she or he won't be there on the set with you teaching you the craft, by word and by example. You won't be able to learn how a film crew works and would never be inspired by their ideas that could only come from someone with a different professional experience and expertise. You'd never build a community trying to crowdsource your locations, props, or extras.
Sure, there are ways for text-to-video tools to save a lot of money for beginning filmmakers. A car crash scene; a spacefight; a T-Rex stomping on Red Square. On the other hand, I remember teaching myself the basics of After Effects and working with a CG supervisor, and that knowledge helped me further on the way. It can be used to create drafts to test one's concepts. But so can a phone camera.
What I ultimately think is that we don’t need billions of dollars poured into thrash video “content”, blackmail fake videos, and thousands more jobs cut to produce what will be mediocre at its best. We need text-to-stop-climate-change, text-to-vaccines, text-to-pay-equality, text-to-end-dictatorships. As they sang in a Soviet kids sci-fi film of my childhood, "robots work hard, humans are happy" — so "robots work hard, and robots are happy" isn't the future I need.
Would you use text-to-video tools in your work? If yes, then how?