Video · Lesson 3.1
12 AI Video Editing Requests You Can Adapt Today
Seeing a good request makes writing your own easy. Below are twelve you can paste and adapt. They are written examples of what to ask, not customer results, and each follows the method in Lesson 3.1.
Be specific about the outcome, name the rule, and ask to see the result before it moves on.
Rough cut requests
1. The basic rough cut
Line up both cameras with the sound. Remove pauses and mistakes. Keep the best take of each line. Show me before you go on.
Why it works: It names the three jobs of a rough cut and builds in a review.
2. The strict keep
Where I said a line more than once, keep the last take only.
Why it works: Gives a clear rule for choosing between takes.
3. Cut for length
Trim this to under 60 seconds without losing the main point. Tell me what you removed.
Why it works: A limit and a report, so you can restore anything important.
Zoom requests
4. Zoom on key words
Zoom in slowly on the key words and back out between them. Check no zoom cuts off my face.
Why it works: Keeps the picture moving and sets a safety check.
5. Something changes every few seconds
Make sure something on screen changes at least every few seconds, using zooms or cuts, not effects.
Why it works: Names the goal, attention, and limits the method.
Colour requests
6. Match two cameras
Make both cameras look like one shoot, warm and clean. Show before and after side by side.
Why it works: One look, with a visual check.
7. Fix a dark face
My face is too dark in the second camera. Lift it to match the first camera without washing out the background.
Why it works: Describes the problem, the target and a limit.
Caption requests
8. Standard captions
Add captions last. Only when I speak. Two lines at most.
Why it works: Three simple rules that avoid clutter and follow final timing.
9. Highlight the key word
In the captions, make the key word of each line stand out in a bold colour. Keep the rest plain.
Why it works: Directs the eye without changing the layout.
Style and learning requests
10. Learn from a reel
Study this reel second by second. Write down how it is edited: shot length, zooms, captions. Save it as my style, then show me what you wrote.
Why it works: Turns a reel into a reusable pattern you can read and correct.
11. Save my fix
I trimmed the intro by hand. Compare my version with yours and save what I changed as rules.
Why it works: This is how edit number two beats edit number one.
12. Save usage
Work from the transcript where you can. Check only a few frames, not the whole video.
Why it works: Usage depends on how long and complex the work is, so lighter inputs help one edit fit in your plan.
How to adapt them
- Change the numbersLength limits and line counts depend on your platform.
- Change the nounsReplace "my face" or "both cameras" with what is in your footage.
- Keep the reviewLeave "show me before you go on" in. It is cheap protection.
- Save what worksAdd the best requests to your template.
Different kinds of video
| Video type | Requests to use |
|---|---|
| Talking tip, short | 1, 4, 8 |
| Longer lesson | 1, 3, 5, 8 |
| Two camera interview | 1, 6, 7 |
| Repeat series | 10, 11 |
Putting several requests together
Single requests are easy to copy. Real sessions chain them. Here is a sequence for one video, using the requests above in order.
- StartUse request 10 once, to teach and save your style.
- CutUse request 1, then review the result.
- ZoomUse request 4, then check your face stays in frame.
- ColourUse request 6, and review before and after.
- CaptionsUse request 8, and add request 9 if you want key words to stand out.
- LearnUse request 11 after you finish your hand fixes.
Request 12 can be added at the start of the session, since it applies to everything.
Common ways to write a request badly
Too vague. "Make it better" gives the assistant nothing to aim at. Name the result.
Too many things at once. Asking for cut, zooms, colour and captions in one message makes mistakes hard to trace. Go one stage at a time.
No review. Leaving out "show me before you go on" removes your chance to catch an error early.
No limit. Without a limit such as "two lines" or "under 60 seconds", the result may not fit your platform.
A note on trust
Every request above asks you to look at the result. That is deliberate. The assistant works fast, and it can be wrong quickly too. A habit of watching the rough cut once, checking a zoom or two, and reading a few captions closely protects you from publishing a mistake. It also teaches you what the assistant tends to get wrong with your footage, which you can then turn into a saved rule.
Questions people ask
Are these real customer examples?
No. They are written examples of what to ask. Use them as starting points and adjust.
Do the requests work with any footage?
They are written for talking videos edited in Premiere Pro.
Where do I put them?
Paste them one stage at a time, in the order: rough cut, zooms, colour, captions.
Where to go next
See requests like these run on screen in Lesson 3.1. Turn the result into shorts with Lesson 3.2, build captions in your style with Lesson 3.4, or fill in the free template.