The Edit Framework: A Repeatable Process for AI Video Editing

Editing by feel gives uneven results. One video comes out sharp, the next feels rushed, and you cannot say why. A fixed order of work fixes that. This post gives you one, based on the method in Lesson 3.1.

We call it the Edit Framework. It has five stages, and each stage depends on the one before.

Rough cut, zoom, colour, caption. Always in that order, because each step follows the timing of the step before it.

The five stages

1. RecordCapture footage that is easy to cut. Sound from the microphone, pause and repeat after a mistake.
2. Rough cutLine up cameras with the sound, remove pauses and mistakes, keep the best take of each line.
3. ZoomAdd zooms on key words so something changes every few seconds. Check no zoom cuts off your face.
4. ColourMake every camera look like one shoot. Review before and after.
5. CaptionAdd captions last, only when you speak, two lines at most. Export.

Add a sixth habit that sits outside the line: Learn. After each video, fix what you disliked by hand and save the change as a rule.

Why this order

Each stage creates something the next one needs.

• The rough cut sets the final length and timing. Anything you add before it will move.
• Zooms attach to words, so they need the final words in place.
• Colour is easier to judge once the cuts are settled.
• Captions follow the exact timing, so they must be last, or you will redo them.

If you reverse any two, you redo work. That is the whole argument for the framework.

What you give the assistant at each stage

The assistant needs different instructions at each step. Keep them short.

A filled in example

Say you filmed a three minute tip about writing better emails, using two cameras.

1. Record. Microphone sound. You tripped twice, paused, and repeated both lines.
2. Rough cut. You ask for a cut. It removes the pauses and the two stumbles and keeps the second take of each line. You watch it once.
3. Zoom. You ask for zooms on the three key words. It checks your face stays in frame. One zoom feels too tight, so you note it.
4. Colour. The side camera looked cooler. You ask for a match. You check the before and after.
5. Caption. You ask for two line captions only when you speak. You export vertical.
6. Learn. You fix the tight zoom by hand, then ask the assistant to compare your version and save the rule: no zoom closer than a set amount.

Make it weekly

A framework only helps if it is easy to repeat. Keep a short checklist beside the editor, run the stages in order, and spend two minutes at the end on the Learn step. After a few videos the rules cover most of what you used to fix.

Adapting the framework to your videos

The five stages stay the same, but what goes inside each stage depends on what you make.

A short talking tip. Keep the rough cut tight, use frequent zooms, and check that captions are easy to read at the size your platform shows them. The lesson describes captions that show only when you speak, two lines at most.

A longer lesson. Fewer zooms, steadier pacing, and a style reel taken from a longer format rather than a quick reel. Save this as a separate style so it does not clash with your short tips.

A two person conversation. The rough cut matters most, because choosing who is on screen is the main job. Be explicit about this in your request.

Common ways the framework breaks

1. Skipping the reviewYou move on from a rough cut you never watched, and a wrong take survives to the end.
2. Mixing stagesAsking for zooms and captions in the same message makes errors hard to trace.
3. Never closing the loopDoing the Learn step only sometimes means the rules never build up.
4. One style for everythingA single saved style tries to cover tips, lessons and interviews and suits none.

Printing it as a checklist

Keep this short list next to you while you edit.

• Record: microphone sound, pause and repeat after a mistake.
• Rough cut: watched once, best takes kept.
• Zoom: key words only, face stays in frame.
• Colour: before and after checked.
• Caption: last, only when speaking, two lines.
• Learn: fixes saved as rules.

A named process feels small, but it removes the daily question of what to do next. That alone makes a weekly habit easier to keep.

Where the framework meets Premiere Pro's tools

Each stage maps to something you can also do by hand. The rough cut can start from a transcript, since Premiere Pro's Text-Based Editing lets you cut text to cut the timeline. Zooms, colour and captions each have their own panels. The framework does not care which route you use, only that you finish each stage before starting the next. Adobe also offers an official Adobe for Claude connector, which covers some Premiere tools but not every feature of the app, so check its page for what it can do. The method in the lesson uses its own setup message.

Why a named process helps

A name makes a process easy to share and easy to remember. If you work with a helper or a colleague, you can say "we are at the colour stage" and everyone knows what has been done and what comes next. It also makes problems easy to locate. When something looks wrong, you ask which stage owns it, and go straight there.

Questions people ask

Can I change the order?

You can, but expect rework. The order exists because each stage depends on the last.

How long should each stage take?

It varies with your footage. Time yourself on the first video and use that as your baseline.

What is the Learn step?

Saving your manual fixes as rules so the next edit starts better.

Where to go next

See the framework running on screen in Lesson 3.1. Add captions in your own style with Lesson 3.4, or turn the finished video into shorts with Lesson 3.2. Track whether it works with the metrics guide.