"Suddenly no one watches YouTube videos?
This matter needs to be sorted out to the root. Nowadays, platform algorithms are getting more and more ruthless, and blind photography alone is no longer enough.
AI can help you make videos, but don't expect it to become a god on its own. I have been in the industry for ten years, and I have seen AI tools that can load several trucks.
What are they doing? Split scripts, find material, and cut films. Sounds amazing?
In actual operation, it is all a trap. For example, if you ask AI to write scripts, it will probably write a bunch of nonsense that no one will read.
Last year, a customer came to me and insisted on using AI to make travel videos. The results are out? It's all templated sentences.
Whoever likes to read the words "deeply empower the local economy".
The truly popular content depends on the sense of reality. AI can help you save time, but you have to do the creativity yourself.
Now there is a new game on the Internet called "AI+ Human Collaboration".
How to do it specifically? Let AI generate the script framework first, and then fill it in yourself.
For example, if you want to make a food video, AI can list the steps, but how to shoot it?
It's up to you to audition with a smelly face and see which expression is the most realistic.
Last year, there was a young man from Singapore who used AI to make a coconut chicken tutorial.
The materials AI found were all old-fashioned, but he simply added dirty shots from his own kitchen.
As a result, the number of views directly increased threefold.
Why? It's real. No matter how smart AI is, it can't beat you with the power you throw a shovel.
Nowadays,"AI dubbing + real-life appearance" is also popular on YouTube.
For example, if you make an evaluation video, AI will give you a calm voice, and you will take a reaction shot yourself.
This mash-up reduces costs and is surprisingly effective.
But there is one big hole to pay attention to: copyright.
The material found by AI may be infringing. Last year, someone was sued for travel videos made using AI.
So when selecting tools, you have to see clearly which materials can be used.
I was recently using a domestic tool called Coze, which has a workflow that automatically lightning protection.
You put in keywords and it will help you choose copyrighted pictures.
But don't expect it to be completely reliable, you still have to check it yourself.
Now YouTube has also introduced new algorithms to suppress duplicate content.
If the AI-generated video is too similar to others, you will be demoted immediately.
So I have to add my own things, such as tablet animation and hand-drawn mirror splitting.
There is a blogger named David. When he does technology reviews, he always likes to draw a small person in the corner of the screen.
Can AI copy his videos? This little move cannot be copied.
By the way, YouTube is recently testing a new feature, and AI automatically generates chapter titles.
You upload a video, and it can be segmented based on the content.
Sounds good? But whether the score is accurate depends on luck.
I tried it, and it always grouped irrelevant content together.
So now the mainstream practice is to write scripts yourself, divide them into chapters.
AI can help you make subtitles. It used to take hours for dictation, but now it takes a few minutes.
But there are so many typos that you have to proofread yourself.
This is what my friend's team does now: AI generates a first draft and then asks interns to revise it three times.
The cost is much lower than outsourcing.
Nowadays in Singapore, many studios use AI to make test videos.
For example, if you want to take a picture of your local life, first use AI to generate dozens of versions to see which data is better.
This method is tough, but it takes a lot of time.
But then again, YouTube is becoming more and more demanding.
In the past, you could just take pictures casually, but now you have to have a cover, title, and description.
AI can help you optimize these. For example, if you enter keywords, it will generate a bunch of title schemes.
But which one? It's up to you to judge.
I suggest reading more titles of similar popular videos to find patterns.
For example, titles with numbers have a high click rate.
AI can help you make video covers. For example, if you give a few keywords, it generates a bunch of pictures.
But don't believe in AI aesthetics. It always likes to generate that kind of space science fiction style.
For local content in Singapore, we still need to take some street views, snacks and so on.
Nowadays,"AI+ Live Broadcast" is also popular on YouTube.
For example, if you perform a live game, AI can automatically edit exciting clips.
But there's a problem: AI doesn't know which moment the audience likes.
I once saw an anchor ask AI to cut a shot that he blocked.
As a result, the audience thought he was joking, and the number of views increased instead.
Therefore, when using AI tools, you must use your brain.
There are many free tutorials online to teach you to use AI to make videos.
But don't believe all the headlines that "generate hot money with one click".
What is really useful are those that talk about specific steps.
For example,"Use this AI tool to generate scripts in three steps."
There was a tutorial that became popular last year, teaching people to write video outlines using ChatGPT.
How to do it specifically? List keywords first and then let AI expand.
For example, if you want to do "Singapore Coffee Culture," ask AI: "Help me list five related topics."
Then you choose one to dig deep.
There is now AI that can help you analyze your competitors.
You enter the channel ID, and it lists the other person's popular videos.
But don't just look at the data, you also have to analyze why it is popular.
I have a customer who used AI to find that his opponent's videos always mentioned "durian".
He followed suit with the durian theme, but it turned over.
Because AI didn't tell you that the other party's fans are all Malaysian.
Therefore, AI data should be viewed in conjunction with common sense.
Now YouTube has a new feature that allows AI to generate trailers.
You upload a few clips, and it automatically compiles a 30-second trailer.
But the effects are uneven, and some are like robots.
I suggest cutting it yourself, even with your mobile phone.
Reality is important.
Be practical. AI tools are changing, but the nature of video remains unchanged.
What do you want to do? What does the audience want to see?
AI just helps you save time, not to help you slack off.
For example, if you want to run a local search shop, AI can find materials, but what does the taste taste?
You have to eat it yourself.
Last year, a blogger used AI to make a local food video.
The material is all web pictures, and the audience can tell at a glance that it is fake.
Therefore, AI is an artifact when used well, but it is a trap when it is not used well.
Many studios in Singapore now do this:
AI makes the first draft and the team polishes it.
For example, let AI write the script, and then let local actors change the lines.
This is fast and true.
By the way, YouTube now has annotation requirements for AI-generated content.
For example, using AI to change faces must be explained in the description.
So you have to read the rules clearly before using tools.
There is now AI online that can help you optimize SEO.
You type in the video title, and it analyzes which words go first.
But don't believe it all, because YouTube algorithms are constantly changing.
I suggest reading more of your peers 'titles to find your feelings.
For example, there are many videos with "2024" now.
AI can help you make video thumbnails.
You give a few elements, and it generates a bunch of solutions.
But don't use too exaggerated, the audience is now aesthetically tired.
For local content in Singapore, we still need to take some street views, snacks and so on.
There is now AI that can help you write video descriptions.
You type in keywords, and it generates a paragraph.
But it is often too long and the audience has no patience to watch.
I suggest writing it yourself and making it short.
AI can help you analyze video data.
For example, the audience will close any clip after watching it.
But don't just look at the data, you have to analyze why.
I once saw a blogger who used AI and found that the bounce rate in the first 10 seconds was high.
As a result, he changed the beginning, and the number of views dropped.
Because AI didn't tell him that the audience liked the slow-paced ones.
Therefore, AI tools should be used with intuition.
Nowadays,"AI dubbing + real-life appearance" is also popular on YouTube.
For example, if you make a popular science video, AI will give you a professional voice, and you will take a reaction shot yourself.
This mash-up reduces costs and is surprisingly effective.
But there is one big hole to pay attention to: copyright.
The material found by AI may be infringing. Last year, someone was sued for travel videos made using AI.
So when selecting tools, you have to see clearly which materials can be used.
I was recently using a domestic tool called Coze, which has a workflow that automatically lightning protection.
You put in keywords and it will help you choose copyrighted pictures.
But don't expect it to be completely reliable, you still have to check it yourself.
Now YouTube has also introduced new algorithms to suppress duplicate content.
If the AI-generated video is too similar to others, you will be demoted immediately.
So I have to add my own things, such as tablet animation and hand-drawn mirror splitting.
There is a blogger named David. When he does technology reviews, he always likes to draw a small person in the corner of the screen.
Can AI copy his videos? This little move cannot be copied.
By the way, YouTube is recently testing a new feature, and AI automatically generates chapter titles.
You upload a video, and it can be segmented based on the content.
Sounds good? But whether the score is accurate depends on luck.
I tried it, and it always grouped irrelevant content together.
So now the mainstream practice is to write scripts yourself, divide them into chapters.
AI can help you make subtitles. It used to take hours for dictation, but now it takes a few minutes.
But there are so many typos that you have to proofread yourself.
This is what my friend's team does now: AI generates a first draft and then asks interns to revise it three times.
The cost is much lower than outsourcing.
Nowadays in Singapore, many studios use AI to make test videos.
For example, if you want to take a picture of your local life, first use AI to generate dozens of versions to see which data is better.
This method is tough, but it takes a lot of time.
But then again, YouTube is becoming more and more demanding.
In the past, you could just take pictures casually, but now you have to have a cover, title, and description.
AI can help you optimize these. For example, if you enter keywords, it will generate a bunch of title schemes.
But which one? It's up to you to judge.
I suggest reading more titles of similar popular videos to find patterns.
For example, titles with numbers have a high click rate.
AI can help you make video covers. For example, if you give a few keywords, it generates a bunch of pictures.
But don't believe in AI aesthetics. It always likes to generate that kind of space science fiction style.
For local content in Singapore, we still need to take some street views, snacks and so on. "
