I’m interested in a craft question that seems more important as AI video improves. A sequence can contain striking individual images and still fail to feel like a film shaped by a director.
For those who have completed AI-assisted scenes or short films, what made the decisive difference: continuity between shots, performance and character consistency, editing rhythm, sound design, a controlled visual grammar, or something else?
I’m especially curious about the transition from exploring possibilities to making deliberate choices. At what point in your workflow do you stop accepting what the model offers and begin imposing a coherent directorial point of view?
For continuity, I think personally it really comes down to two things:
Script
Continuing from previous generation
And as for directing, I consider it this way, I am the writer and even the director, so I can decide the lines, the pacing, even dictate how I want the shot to be framed, and camera movement. But the AI is the actual cinematographer and does the camera work. I do not have the ultimate control, which is for me fine, because at times you discover that an AI that has been trained with millions of videos, might actually know better how to film a scene.