Wait until we have multilayered AI models. One to generate the 3d scene. One for the lighting. Three for cameras. One for each actor acting out the given scrip

Wait until we have multilayered AI models. One to generate the 3d scene. One for the lighting. Three for cameras. One for each actor acting out the given script along with a set amount of "improvising" and "randomness", one to generate background royalty free music based on the emotional context of the scene, one to serve as a Foley artist for environment sounds. The equivalent computer power of a medium sized bitcoin farm running everything with the only humans being the writers, the IT people running the systems, a director watching each take as it is rendered live to pick the best ones, and an editor to slap all of the takes together. That's probably technically possible now, if not actually implemented. Another 15 years or so and kids will be rendering full movies in the basement and watching them with their friends.