Generative AI lets me make films from my desk instead of 7 AM call times, steep bills at the rental houses, and unreliable post editors who fail to meet deadlines.
This isn't just a net win, this puts power in my hand I've never had before.
It's still work and art. No LLM is going to make a compelling story or make the right artistic choices. I do that. But now I get to control way more myself and I'm empowered to see the entire vision though.
I always felt that the beauty of film is in the collaboration between all the people involved to create something larger than themselves. Everybody in the credits brings something irreplaceable to a film, and together they transcend a singular vision to build a work of art.
This generative stuff feels reductive to me. There are no actors, no set designers, prop masters, musicians. Nobody’s bouncing ideas off a colleague or working with three other departments to bring a scene together. It feels less like art when it’s a computer using models based on real stolen art to generate content off of a prompt.
You are not, the LLM makes bulk of work for you and will choose a lot of things for you for the movie you are making.
>I'm empowered to see the entire vision though.
I think you fail to grasp one important thing about art in general - it is non-verbal by it's nature. You can't go and explain in LLM input some famous painting, it not how it works.
My partner is a graphic designer and now uses a lot of generative AI in her work. It’s very much an iterative process. She would run hundreds (sometimes thousands) of prompts over the course of compositing a single image. There’s huge amounts of editing involved, too. It doesn’t take less time than before when she was primarily an illustrator. But it enables her to do different types of artwork she wasn’t previously able to.
It’s definitely different. And has some bad sides for sure. But professionals using LLMs for creative work tends to be a lot more involved than just typing a prompt.
That's like how LLMs help me with software development. I don't work less, instead I produce more and what I produce is of greater benefit to my clients.
Diffusion models aren't LLMs and aren't necessarily text promoted. You can paint with them.
> You can't go and explain in LLM input some famous painting, it not how it works.
When directing a film, you're issuing verbal commands to your team. It's actually quite similar to prompting. And I almost never get what I envision. Diffusion in a way gets me closer to what's in my head.
I thought that was one of the problems with some of the art models - that you could input the right sequence of words and get exact copies of famous copyrighted images out.
This isn't just a net win, this puts power in my hand I've never had before.
It's still work and art. No LLM is going to make a compelling story or make the right artistic choices. I do that. But now I get to control way more myself and I'm empowered to see the entire vision though.