AI video moved so fast that most of us never caught up to what it can already do. The ads that used to mean high-end production now sit at the fingertips of any talented creator. The limits are real, but they are shrinking, and they are not the point. The human is still at the center.
How I think about this from inside the work, after about fifteen years making video and now building Tonpit.
The era of the editor
Strip away the tooling and the human is still the source: the idea, the central motion that drives the whole thing, and the hand that ties it back together in the edit. I would go further and call this the era of the editor. The director of the edit.
What changes is the skill set. Editing always asked for a certain emotional and rhythmic intelligence, and, until now, a stack of technical ability to go with it. With AI in the loop, the technical part becomes a useful bonus, not a requirement. The editor turns into a director. A director with the rare ability to run a shoot day from the comfort of home.
A director who can run a whole shoot day from his living room.
The limits, and how you beat them
The limits are real, and naming them precisely is how you get past them. Four show up most, and they share a tell: the result reads as off before you can say why. That is the uncanny feeling, and it is usually one of these.
There is a reason these cluster. A model predicts frames probabilistically. It holds no persistent sense of who a character is or how weight and contact behave, and it optimizes each clip on its own rather than the arc of the whole piece. That is why performance drifts, why physics wobbles, and why it defaults to safe, shallow coverage.
And notice the fix kept repeating. Performance, physics, proportion, language: the answer each time was the same, cinematic language and editing judgment. Every one of these limits resolves into a single human skill. That is not a coincidence. It is the whole argument.
The gaps will close. The edge moves up.
I will be honest about the direction. I would not bet against any of these limits closing. I think AI will likely close all of them.
That does not shrink the human role. It relocates it. The person is still the one who brings the reason, the need, and the purpose the work is meant to serve. The steering stays human. It just rises a layer. As the technical floor falls away, what is left is judgment, taste, and knowing why you are making this at all.
The tools will close the gaps. The reason to make something stays ours.
Two short follow-ups coming on the crafts that do most of the heavy lifting here: holding consistency with image references, and cinematic language itself.