"Is your account still alive?
Want to use AI to make YouTube videos? Don't worry, figure out a few things first. Nowadays, many people think that AI can generate hot money with one click, but as a result, the video is posted and no one watches it. Why? Because machines don't understand people.
Let's start with the most practical. What can AI help you? Write scripts, generate pictures, and cut films. For example, use Midjourney to compose pictures, use Runway to carve pictures, and use Picture to edit automatically. These tools are really easy. But there's a big problem-the machine doesn't know who your audience is.
I have seen many novices using AI to generate a bunch of "perfect" videos, and the results are played in single digits. Why? They only think about technology and forget the content. YouTube is not a Short Video platform. Users want information density. Scripts written by AI are often full of nonsense and the pictures are jumping around.
You have to prioritize. AI is a tool, not a brain. For example, when doing technology evaluation videos, AI can help you find information and write outlines. But how to put it concretely, it's up to you to keep suspense and where to interact. Last year, I helped a client make AI videos. He asked the machine to write scripts, but the audience said,"It's like listening to instructions." Later, we manually changed it three times, and with the addition of real people appearing in the audience, the number of views immediately increased fivefold.
Several popular AI tools now have their own pitfalls. For example, Runway's Gen-2 generates pictures quickly, but the logic often goes wrong. If you let it take pictures of "people walking in the rain", it may give you a sunny day. So test with simple scenes first, and don't complicate the plot.
There is also dubbing. AI dubbing saves time, but the listener can hear it. ElevenLabs, for example, sounds like reading a script. I suggest using real people to dub key parts and using AI to supplement others. A blogger made a popular science video last year. The first half of the video was given by real people, and the second half was read by AI. Audience feedback said that "the first half was like chatting with friends, and the second half was like broadcasting."
Don't be fooled by the saying that "AI can replace editors". Machines can do basic editing, but people still have to rely on emotional rhythm and picture transition. For example, last year's YouTube algorithm update paid more attention to user stay time. If you don't catch anyone in the first five seconds of your video, no matter how AI optimizes the middle section, it won't work.
Practical advice: Use ChatGPT to write the outline first, and then adjust it manually. For example, the theme "How to use AI to make YouTube videos" can be broken down into: selecting tools, writing scripts, taking materials, and cutting videos. Find AI-generated material for each part and then manually integrate it. I had a student who did this. He made 20 videos a month, and the average number of views was 30% higher than that generated by pure AI.
Be honest. AI tools are changing, but audience tastes remain the same. What they want is "useful" and "interesting". For example, if you make a travel video, AI can generate scenery pictures, but the audience prefers to see how local people live. More than 60% of popular YouTube videos last year were live-action.
So don't use AI as a panacea. It is suitable for handling repetitive tasks, such as finding music and adjusting colors. The core creativity still depends on you. For example, if you take a tutorial video, AI can help you write the steps, but how to explain it more clearly depends on you. I have seen a food blogger who used AI to generate recipe videos, but the audience said that "the steps cannot be understood." Later, he demonstrated it in person, coupled with small error demonstrations, and the number of views doubled.
Now YouTube searches for "AI video production", and the results are full of tutorials. But few are really useful. For example, many people recommend using Synthesia as virtual humans, but forget to mention that the audience is tired of watching it. In the second half of last year, the average completion rate of virtual human videos was 20% lower than that of real people.
To sum up: AI is an assistant, not a brain. Think clearly about what the audience wants first, and then find tools. For example, when doing knowledge videos, AI will help you find information; when doing life videos, AI will help you find music. At last year's YouTube creators conference, Google said there would be more video tools in the future, but the core is still content.
So don't rush to fully AI. Use it first to solve the most annoying parts, such as subtitles and dubbing. For example, I recently used Descript to convert recordings to subtitles much faster. But key parts, such as the first golden three seconds, you still have to do it yourself. Last year, a channel used AI to generate all videos, but viewers found that "all videos started the same" and complained to YouTube about streaming restrictions.
A piece of advice: YouTube recommendation algorithms now value interaction more. AI can generate comment responses, but viewers can tell whether they are true or false. For example, if you ask AI to reply to a "great video", the audience will think it's fake. Last year, a channel used AI to reply, but was reported and punished by the algorithm.
So use AI when you need to, but don't be lazy. For example, use AI to generate topic selection, but you still have to think about the content yourself. Last year, I helped clients make AI videos and found a problem: machine-generated topics were either too broad or too unpopular. Later, we screened it manually and the effect was good.
Now when doing YouTube, AI is a helper. For example, use Canva to design titles, use Picture to automatically edit, and use ChatGPT to write scripts. But you still have to control the key points yourself. For example, last year, YouTube updated its AI content tags, and machine-generated content must be actively labeled. Viewers now trust videos made by real people more.
So don't be scared by the words "AI replaces humans". Last year, I helped a client make AI videos. He wanted to fully automate it at first, but found that what the audience wanted was reality. Later, when we appeared in real person and added AI assistance, the effect was good.
Be honest. Nowadays, there are many AI videos on YouTube, and viewers are tired of watching them. For example, virtual anchors and AI dubbing were still fresh last year, but this year the audience's response was flat. So don't blindly follow the trend, think about what your audience wants first. For example, when doing tutorial videos, AI can help you write the steps, but it depends on you how to explain them more clearly. Last year, a technology channel used AI to generate all videos, but the audience said "I couldn't understand it." Later, when they appeared in real life and added close-ups of their hands, the number of views increased.
So AI is a tool, not a brain. For example, use AI to find topics, but you still have to think about the content yourself. Last year, I helped clients make AI videos and found a problem: machine-generated topics were either too broad or too unpopular. Later, we screened it manually and the effect was good.
Now when doing YouTube, AI can help you a lot. For example, use Runway to generate pictures, use ChatGPT to write scripts, and use Descript to cut slices. But the core idea still depends on you. For example, more than 60% of popular YouTube videos last year were live-action.
So don't use AI as a panacea. It is suitable for handling repetitive tasks, such as finding music and adjusting colors. The core creativity still depends on you. For example, if you take a tutorial video, AI can help you write the steps, but how to explain it more clearly depends on you. Last year, a food blogger used AI to generate a recipe video, but the audience said that "the steps couldn't be understood." Later, he demonstrated it in person, coupled with small error demonstrations, and the number of views doubled.
Now YouTube searches for "AI video production", and the results are full of tutorials. But few are really useful. For example, many people recommend using Synthesia as virtual humans, but forget to mention that the audience is tired of watching it. In the second half of last year, the average completion rate of virtual human videos was 20% lower than that of real people.
So AI is an assistant, not a brain. Think clearly about what the audience wants first, and then find tools. For example, when doing knowledge videos, AI will help you find information; when doing life videos, AI will help you find music. At last year's YouTube creators conference, Google said there would be more video tools in the future, but the core is still content. "
