The first lesson for short video novices: I wrote the editing into a program, and then I realized that all the skills that need to be practiced are before pressing the record button.

At the end of every workshop, someone almost always stays and asks me the same question: Which microphone should I buy first? Do you want fill light? Should the mobile phone be replaced with a new one?

I usually ask back: How many have you taken?

The answer is often not yet. It sounds reasonable to have the equipment ready before starting.

Two months ago, I wrote a manuscript about this matter. The argument was very clear: don’t chase the equipment, shoot first and then talk about it. The real obstacle for novices is not technology, but taking preparation as the beginning.

Then the manuscript sat in the draft box, untouched for sixty days.

I later understood why. That article says all the right things, but it only answers half the question. It tells you to stop waiting and shoot, but it doesn’t tell you what you need to do to get better after shooting. And this stretch of road in the middle is exactly where most people disappear while walking.

What really made me see what was missing was when I later wrote down my entire editing process into a set of programs.

Writing clips into a program was the first thing I encountered

In mid-July, I did something that had nothing to do with writing: I no longer used editing software to edit the film. Instead, I wrote down rules that a computer can understand on how to cut well, and gave them to the machine to check.

The most honest thing about programming is that it doesn’t accept adjectives. You can’t tell the computer that the rhythm of this film should be better. You have to explain clearly: What does the rhythm of this film mean? At what second or over how many seconds does it count?

The first version was up and running in half a day, but for more than a week, I kept adding rules to it. In the end, more than 200 tests were guarding these rules. What really surprised me was how long the forced list was:

Each of the four positions of 0%, 25%, 50%, and 75% of a film must have a stimulation point. If one is missing, a warning will appear, because those are the four critical points where the audience is most likely to lose focus.

As long as the total length of the supplementary screen is equal to the main screen, it will be directly judged as a failure and will not be given. Because that means the supporting character steals the show from the protagonist.

After the call to action at the end of the film is finished, it doesn’t end for more than fifteen seconds, warning. Tailing is very detrimental to the completion rate.

Within ten seconds before the end of the video, as long as there is silence for more than five seconds, the judgment will fail. This rule left a sentence in the code as an error message: “The sound cannot be cut off before the picture is played.”

There is another one that I find most interesting. The system will read the word-by-word time code of speech recognition and find the pauses between two adjacent words in the same sentence. Anything longer than 0.6 seconds is marked as stuck. Note that it is within the same sentence, not between sentences. Because it’s natural to pause between sentences, if you get stuck in the middle of a sentence, the audience slips away.

After I finished writing these rules, I looked back at the list and was stunned for a moment.

None of this is talent.

None of them require beauty, aura, or your natural camera sense. Each item can be measured, checked, and made into a list and followed accordingly.

And I’m almost certain that the people who queued up after the workshop to ask me which microphone they should buy were never given this list.

Between labor and judgment, there is still a third layer stuck

I have said before that a film actually has two levels: one is labor, which is aligning the materials, typing the subtitles sentence by sentence, and turning the volume to just right; the other is judging which paragraph is the most touching and whether the copywriting speaks to the heart. I said in “The threshold of audio and video has not disappeared, it has just moved” that AI takes over the first layer and leaves the second layer intact to you.

I still agree with this statement, but it misses a piece.

After writing the editing rules into a program, I realized that there was still a third layer stuck between labor and judgment. Let me call that layer specifications.

Specifications neither consume your energy nor test your taste. It only tests one thing: do you know.

You don’t know that there can’t be five seconds of silence at the end, so your film ends abruptly at the most exciting point. The audience finds it weird but can’t tell why. You don’t know that if you are stuck in the middle of the same sentence for 0.6 seconds, the person will fall off, so you think in the middle of the sentence when recording, and it sounds very awkward. You didn’t know that you couldn’t have an empty scene in the first three seconds, so you carefully made the opening animation to just keep everyone out.

These are not labor issues, and you can’t fix them no matter how hard you drag the timeline. It’s not a matter of judgment, you probably have better taste than a lot of people. It’s just a matter of knowing or not knowing.

Ninety percent of the most frustrating part for novices falls on this level. But almost everyone misjudged it as a second-level problem, and concluded: I probably don’t have this talent.

No. You’re just locked out with a spec sheet that no one sent you.

The specification layer has a very good feature: it does not need to be practiced, as long as someone tells you. The taste depends on the vintage, and the specification is something you can use on your next piece after hearing about it. When comparing the three layers, the difference is very clear:

Level	What will test you	How to make up for it	Who should do it
Labor	Physical strength and patience	No need to supplement	Leave it to the machine
Specifications	Do you know	Someone gives you a list	You will know it after reading it
Judgment	Opinions and tastes	Can only be accumulated over time	It can only be you

Most novices spend their energy on labor and put their anxiety on judgment, but no one ever tells them that there is a layer of specifications sandwiched in between.

The most disturbing thing is that half of this list should be used before shooting starts

If the story stopped here, the conclusion would be: novices should learn a post-production checklist.

But after I finished my first real-life video, I discovered something much more disturbing than this.

Let me make one thing clear first: that piece is the essence of the horizontal class. Of course I have cut straight short videos, and more than one. From mid-July to late July alone, I delivered more than a dozen 1080×1920 finished products, including course highlights, word-of-mouth short films, and workshop highlights. But none of them were cut by that system. The system’s straight rules are still on the to-do list to this day, and I made the dozen or so of them by hand-carving pipelines separately; three of them were to first let the system produce horizontal master tapes, and then write a script to rearrange the images into straight format.

This gap itself shows one thing: I have handed over the labor layer to the machine, but I have not yet programmed the specifications that the straight line should abide by, so that part still has to be watched by humans. And the things I noticed happened to fall on the shooting side.

Back to that classroom gem. There are only four files of material, which add up to more than two minutes, and were finally cut to eighty-five seconds. For such a short section, the editing decision went through six versions.

The first version is a combination of four essences, which lasts about ninety seconds.

In the second version, the only thing I did was cut out the redundant words at the beginning. The original words are “It’s equal to”. After cutting out, the length of the film became 87.74 seconds.

The third version is the same, with the hanging “and then there is” at the end cut out, and the length of the film is 85.74 seconds.

Two rounds of so-called fine editing saved a total of more than four seconds, and all of them were spent on deleting nonsense. There were no transitions, no special effects, nothing to make the film look more powerful. To be honest, when it comes to editing, most of the time it’s about subtraction.

And these four seconds didn’t need to be cut at all. If I didn’t say those two mantras when recording, these two rounds of processes would not exist.

The more cruel ones are yet to come.

When the filming of that material started, the camera was still moving and shaking, and people had already started talking. I mentioned this once in “I let AI take over the editing”. The focus at that time was the blind spot of the system. What I want to talk about this time is its accounting: the first time I tried to remedy the situation, I used another frame to cover the first 6.4 seconds of the opening, which was not enough; when I looked back, I found that the footage didn’t really stabilize until the eleventh second, and in the end I had to cover it all the way from 0 seconds to 12.34 seconds.

For an eighty-five-second film, the opening twelve seconds are made up of other scenes. The reason is that when I started shooting, I didn’t wait for the camera to stop.

I wrote the root causes into the to-do list and recorded them as blind spots in the system. But the one piece of advice that is actually left in the user manual and written for people to read is this: “After turning on the phone, let the camera stabilize for 2 to 3 seconds before starting to talk.”

After going full circle, the first piece of advice given to people by a system that automatically edits videos is to be quiet for three seconds after pressing record.

No need to spend money, no need to learn software, no need to wait for AI to improve. Just three seconds.

By the way, this 85-second film actually took two hours from the material landing to the final film, and there were more than a dozen intermediate versions in the meantime. I want to leave this number to everyone who feels that their first shot was terrible: a person who has a whole set of automated processes in his hands, processes eighty-five seconds of footage, and has to go back and forth for six revisions before calling it a day. You re-recorded the first song three times, which is very normal.

Straight taught me two more things

When I was cutting the dozen or so straight cuts, I was bitten twice more by the specifications, and the biting positions were very representative.

The first time is that the same set of parameters behave completely differently in two formats. My hand-carved pipeline uses an effect that slowly pushes the static image closer. The horizontal position looks normal. When the same parameters are changed to the vertical position, the person becomes taller and thinner. The reason is that when zooming in, I directly pulled the picture to the target size without locking the ratio; the 16:9 canvas cannot be seen clearly, and the details are revealed as soon as I zoom in to 9:16. The solution is to enlarge the image first and then crop the excess part. I still remember the lesson from this incident: if you want to test, just test the one with the farthest difference in specifications. Don’t test just one and assume they are all correct.

The second time is more worthy of writing for those who haven’t started filming yet. There is a batch of materials that I want to cut from horizontal format to vertical format. Before doing so, I first measured the position of the speaker in the screen. As a result, the horizontal center of the face in the five paragraphs fell at 38.0%, 44.1%, 44.9%, 45.6% and 47.2% respectively, all to the left. In other words, if you follow your intuition and cut it from the center, each section will cut off a piece of the person.

This discovery didn’t turn out to be some clever automatic cropping algorithm. It became the last item on the list.

A checklist to use before you press the record button

Putting all the above back together, I rewrote the table for newbies. It’s not in the editing stage, it’s in the shooting stage.

First, press record and shut up for three seconds. Stop the camera and let the radio noise subside. There is also a side benefit of these three seconds: during post-production, there is always a clean frame at the beginning that can be used as a buffer. This is the one with the highest reporting rate in the entire list.

Second, get right to the point in the first three seconds. No empty scenes, no opening animations, and no introduction of yourself by name. After all, your first three seconds are for the audience to decide whether to stay, not for you to warm up.

Third, stop after finishing a sentence. If you want to think about something, think in the gaps between sentences, don’t stop in the middle of the sentence. If you pause for more than 0.6 seconds in the same sentence, the listener will feel stuck. This was something that was very troublesome to fix in the editing stage, but almost effortless to change in the moment of the speech.

Fourth, after finishing the last sentence, don’t rush to close it. Record for three more seconds before stopping. The hard cut at the end is one of the most common flaws for beginners, and its cost is those three seconds.

Fifth, read your verbatim draft first after shooting, don’t look at the picture first. The first step of my system was designed like this: just look at the text to decide which words should be kept and which words should be cut, and the pictures should be completely dealt with later. If you read it yourself, you will realize how dense your catchphrases are and how long your nonsense is. This step is more useful than watching any lens movement tutorial.

Sixth, if you want to send it horizontally or vertically, shoot the same paragraph once horizontally and again vertically. This is the easiest way to do it. If you expect post-production to cut the horizontal ones into straight ones, you will most likely find that the person is not in the center.

These six items have one thing in common: they all cost zero, have zero equipment, and can be done immediately. None of them were thought up by me. I wrote the editing rules into a program, and then pushed them back after being slapped in the face several times by the films I edited.

To use an analogy, it’s like cooking. No matter how good your knife skills are, you can’t save a vegetable you bought by mistake. Post-production is knife skills, and these six items are food selection and ingredient preparation, which happen before you enter the kitchen.

Those things that the machine cannot catch are the ones you have to practice for ten years.

At this point, I need to explain everything completely, lest you think that just following the list will make a good film.

When I designed the film review mechanism, I set a rule for myself: the subjective quality of the film will never be given a comprehensive score, but only a red light, a yellow light, a green light, and a one-sentence explanation. The reason is written in the document to avoid false accuracy on subjective things.

Whether a film looks good or not, there is no way it can be rated 87 points. If anyone gives you this score, the decimal point is fake.

When sorting out the methodology, I also marked each item with a category. One category is called manual judgment points, which refers to those that require people to make aesthetic or content judgments and cannot be written according to rules. There are more entries like this than I originally expected. The stimulation point should be placed at 25%, which is the specification; what counts as the stimulation point is judgment. There must be a hook in the opening, which is a specification; which sentence should be said on that hook, which is a judgment.

There is another thing that I think should be told to novices: the traffic limit of a video is actually determined for the most part before you start shooting. It is determined by how many people care about the topic you choose, and then it is the performance and editing.

In other words, topic selection is more important than camera movement. Of course camera movement is useful, but the question first marks the ceiling, and editing determines how close you are to that ceiling. Therefore, the order of practice should be as follows: on the level of specifications, once you know it, don’t spend any more energy on it; all the remaining energy should be spent on the level of judgment. Someone who reverses the order will run a technically clean channel that no one wants to watch.

So, I changed the conclusion I made two months ago

What I wrote two months ago was: Take 30 pictures first, and you are not allowed to delete them, and you are not allowed to worry about whether they are good or bad.

Now I will change it like this.

Just shoot ten shots first, but before shooting each one, run through the six items above in your mind. The meaning of thirty shots is to master a process, but if you shoot all thirty shots with the same set of bad habits, you have mastered bad habits.

After filming, do only one thing: read your verbatim script and make a list of your mantras. Mine is “is equal to” and “and then there is”, yours must be different. Make a list and stick it next to the screen so you can take a look before shooting the next one. The effect of this thing is better than buying any microphone.

There is another exercise that I highly recommend, and the cost is also zero. Pick a movie that you think has a good rhythm. While watching it, pause every section and write down two things: What are you watching now? What’s your mood now? Draw these dots on a circle. After taking apart a few more, you will find that there is a pattern in where good films are placed, and that pattern will slowly grow into your intuition. It’s taking someone else’s taste and breaking it down to your own specifications without having to wait for anyone to teach you.

Back to the workshop question.

The next time someone asks me which microphone to buy, my answer will be different – ​​the real answer is three seconds. After pressing record, be quiet for three seconds to let the camera freeze, and then go straight to the key points. Stop after finishing one sentence. Don’t rush to turn off the video after the last sentence. that’s all. With that old cell phone in your hand, coupled with these tips, you can already take better photos than half the people.

Don’t chase the equipment. I still stand by this statement today.

But two months later, I will put it more accurately: What you lack is never the equipment, but a specification sheet that no one sends you. And the most useful items on that list are all available before you press the record button, and you don’t have to spend a dime.