The Taste Rule

Somewhere in the middle of every conversation about AI editing, someone asks the real question: "Okay, but where's the line?" Here's mine, and it's the spine of this entire course. Every task in post is either labor or judgment. Labor is anything where a correct answer exists: is this clip in sync, what words were spoken, does this shot match that shot's exposure. Judgment is anything where the answer depends on what the film is trying to make someone feel. Labor gets automated the moment a tool can do it. Judgment never gets handed over — not because of principle, but because the tools are genuinely bad at it, and I'll show you why.

The taste rule sounds obvious written down. In practice, the tools blur it constantly — an assembly assist that also trims your pauses, a "smart cut" that decides pacing for you. So let's draw the line item by item.

Automate this. Never that.

• Automate: transcription and logging. There is a correct transcript. Let the machine type it.
• Automate: sync, grouping, and organization. Matching audio to picture and sorting takes by scene is arithmetic wearing a media browser.
• Automate: technical fixes and exports. Noise floors, format conversions, caption files, platform deliverables — correct answers exist for all of them.
• Never: shot selection for performance. Which take is alive is not a measurable property. It's the job.
• Never: the length of a pause. A trimmed silence and a devastating silence look identical to an algorithm.
• Never: what to cut for pace — and never the ending. Pace is a promise you're making to the audience, and the ending is the whole reason the film exists.

Notice the pattern. Everything on the automate side has a verifiable right answer. Everything on the never side is a bet about human feeling. The moment you can't say what "correct" would even mean, you've crossed into judgment — and the machine is now guessing on your behalf.

Why AI cuts feel dead

Here's the honest mechanics of it. When an AI assembles a scene, it's drawing on patterns — what cuts usually look like, how long shots usually hold, where scenes usually start. The result is a cut that resembles the statistical center of every cut it has seen. That's why AI edits have a specific, recognizable quality: nothing is wrong, and nothing is true. Competent and dead. The take it picks is the clean one, not the cracked one. The pause it leaves is the average pause, not the one that lets a line land. Films work by deviating from the expected at exactly the right moments — and deviation from average is, by definition, the thing an average-seeking system edits out.

So has the editor's job changed? Read any honest description of it — choose the performance, control the rhythm, protect the story, know what the audience feels at every second — and notice that none of those line items has moved an inch. The job description hasn't changed. The tools around it have. What's changed is how much of your week goes to the job versus the chores orbiting it. Editors who understand that are about to do the best work of their lives, because the drudge work that used to eat half the schedule is evaporating. Editors who hand the judgment over too will ship films that feel like nobody made them.

That's the foundation module complete. Next we get our hands dirty with the single most useful AI skill in post: transcription, and the strange new craft of cutting video like a document.