What does film editing teach that transfers to AI filmmaking?

Editing teaches pacing, continuity, coverage, and taste — judgment about which shots belong together and how long each should hold. Those instincts decide whether generated footage becomes a film or stays a reel of clips. The tools changed; the craft of assembly did not.

Is editorial experience actually useful for AI-generated video?

Yes. AI models produce clips; someone still has to make them cut. Knowing what reads on screen, where a scene needs coverage, and how long a moment plays is exactly the value a model does not provide. That editorial judgment is the difference between a piece and a demo.

Why does coverage matter when you can generate any shot?

Because generating a shot and having the right shot are different things. Editorial thinking tells you in advance what angles a scene needs to cut cleanly, so you generate coverage on purpose instead of discovering a gap in the timeline with nothing to fill it.

Does AI replace the film editor?

No. AI changes what an editor works with, not what an editor decides. Pacing, continuity, and taste still live with a person. If anything, generation makes editorial judgment more valuable, because it is now the main thing standing between raw output and a finished film.

← Writing/January 13, 2026·3 min read

From Marvel Edit Bay to Neural Net: What Editorial Craft Teaches AI Filmmaking

What ten years of editing at Marvel Studios taught me that transfers directly to AI filmmaking — pacing, continuity, coverage, and the taste to know what actually cuts together.

By Wayan Palmieri

A decade in the Marvel Studios edit bay taught me that filmmaking lives in the assembly, not the capture — and that lesson transfers to AI filmmaking untouched. Pacing, continuity, coverage, and taste decide whether footage becomes a film or stays a pile of clips. AI changed where the clips come from. It changed nothing about what makes them cut.

I spent roughly ten years in-house at Marvel Studios, cutting across twenty-five feature films, two shows, and more than fifty trailers, and working as a 3D stereoscopic editor. I am a member of the Motion Picture Editors Guild, Local 700. I now direct AI production and make generative films of my own. The through-line between those two lives is the whole point of this essay.

What a decade in the edit bay actually teaches

You do not learn editing from the buttons. You learn it from thousands of hours deciding what stays, what goes, and how long a moment holds before it dies on screen.

That work drills in a handful of instincts you stop being able to switch off:

Pacing — feeling the exact frame a shot has given everything it has.
Continuity — catching the eyeline, the direction, the light that will not match before an audience does.
Coverage — knowing what a scene needs to cut cleanly, and noticing the gap early.
Taste — the unglamorous judgment about what is actually good, which no tool hands you.

None of that is about a specific NLE. It is about how images read in sequence to a human being.

Why those instincts transfer straight to AI

Generative models are extraordinary at producing a clip. They have no opinion about whether that clip belongs in your film.

That is the gap editorial fills. A model will happily give you a gorgeous shot that runs two seconds too long, breaks the eyeline from the shot before it, or covers a scene from an angle that cannot cut. The model does not know. The editor does. Everything I learned about assembly is exactly the layer generation does not provide.

The most concrete version: coverage. On a traditional show you plan angles so a scene can be built. In generation you can make any angle you want — which means you have to decide what to make, in advance, with an editor's sense of what the cut will need. The freedom is total, so the judgment matters more, not less.

Edit bay vs neural net: what changed, what did not

Element	Marvel edit bay	AI filmmaking	Same or changed
Source of footage	Shot on set	Generated by models	Changed
Pacing	Editor's call	Editor's call	Same
Continuity	Must match	Must match	Same
Coverage	Planned on the day	Generated on purpose	Same discipline, new mechanism
Taste	The whole job	The whole job	Same
Iteration speed	Bound by shoots	Bound by compute	Changed

The left three columns of that table are the ones that decide the film. They did not move.

The craft is the moat

There is a lot of talk about AI replacing filmmakers. From inside both worlds, that gets it backwards. Generation floods you with footage; the scarce thing becomes the judgment to shape it. When anyone can make a clip, the person who knows what cuts together is worth more, not less.

This is the same argument I make about where AI-hybrid and AI-native production actually differ — not the tools, but the discipline behind them. It is also why an AI-native pipeline lives or dies on its editorial stage.

I will admit a bias here. My great-grandfather, Hans Bartholowsky, painted mattes in Hollywood's Golden Era — inventing images that were never photographed, long before anyone typed a prompt. The tools have always changed. The craft of making them believable never has.

If you are building AI production and want that editorial instinct in the room, I consult on pipelines and creative direction. For films made this way, see selected AI Artist work.