The line that gets repeated about AI filmmaking is that one person can now make a whole film. It is worth being precise about how true that is, because the gap between the slogan and the reality is the actual story, and it leads straight into the harder question of who counts as the author.
At feature scale, the honest number is small, not one. Critterz, the animated feature made with Sora, was built by a modest team in about nine months for under thirty million dollars, against the hundreds of people and the years a traditional animated feature takes. The compression is real and dramatic, an order of magnitude fewer people and months instead of years, but it is compression, not disappearance. Where the team of one genuinely exists is at smaller scale, and it is remarkable there: a single ad director delivering broadcast-grade spots in days, or Neural Viz sustaining an entire fictional television universe largely alone.
The through-line is that AI collapses the crew, not the craft. It removes the people whose job was execution, the render farm, the camera crew, the room of in-betweeners, and leaves the people whose job is judgment, the writer, the director, the editor with taste. That is why the winners so far are not prompt typists but experienced filmmakers and operators who know what they are looking at. And it is why the tools reward moving up into direction rather than staying at the console.
So who made the film. That question stopped being academic in 2026, and it came to a head at the festivals. The most prestigious venues in cinema have moved to keep fully generated work out of their main competitions even as AI tools are used at the edges of programs and technology companies take festival partnerships. The specifics are contested and still settling, so treat any precise account of the rules with caution, but the direction is unmistakable. The gatekeepers are drawing a line, and for now the line puts fully generated work outside the door.
The more useful framing comes from the working side of the argument, and it lands close to this book's thesis. Creativity did not disappear, it moved upstream. When execution is cheap and automatic, the scarce and valuable act is the one that decides what to execute: the taste, the story, the judgment about which of a thousand possible shots is the right one. Prompting is a tactic. Direction is a discipline. The tools got good enough that having something to say became the bottleneck again.
The authorship question stopped being theoretical this cycle and now has a legal floor. The US Copyright Office's January 2025 report holds that purely AI-generated material is not protectable and that prompts alone do not establish the human authorship copyright requires, while human-authored elements, arrangement, and substantial modification of outputs can be. In March 2026 the Supreme Court declined to review Thaler v. Perlmutter, leaving intact the rule that copyright requires a human author. So who is the author now has an answer for legal purposes: the human whose creative control is substantial and demonstrable, not the person who typed the best prompt. That is also why studios building for release invest in provenance tooling that logs prompts, settings, and approvals, because a film that cannot secure a registration or clear its likenesses is hard to sell to a streamer or insure for theaters.