"Your video dubbing suddenly becomes a real person?
Recently, many people have discovered that their TikTok accounts are becoming more and more like real people. There are actually two black technologies hidden behind this: artificial intelligence speech generation and video production tools. They are now joined together, making ordinary people "video producers."
Let's talk about speech generation first. In the past, AI dubbing was always like a robot reading manuscripts, but now it is different. For example, iFlytek's new technology can imitate hundreds of sounds, and even modal particles can be learned well. Last week, I also helped my client test it, and even the coughing sound was added to the manuscript it read.
Video production tools are even more powerful. Platforms like coze package editing, special effects, and subtitles. You just type text and it generates video. Last week, my friend used this to make a travel Vlog. The film was released in five minutes, which was much faster than when he cut it manually.
Why is it so popular now? Because the threshold is too low. In the past, you had to learn editing software to make videos, but now you just need to know how to type. I have a customer who has never touched videos at all. Last week, he used this tool to make a product introduction, and the number of views exceeded 10,000.
But pay attention to a few pits. First, copyright issues. The sounds and pictures generated by AI may involve authorization. For example, the dubbing of DeepVoice depends on whether it is commercial. Last week, a blogger was warned by the platform for not paying attention.
Secondly, homogenization is serious. Many people now use ready-made templates, and videos all look the same. Last week, I swiped a food video and found that three accounts used the same set of AI to generate images. The audience will get tired of watching it too much.
So how to make differentiation? Try these methods. First, mixed real people appear on the scene. For example, AI provides a background introduction, and you tell the key points. I have helped customers do this, and the completion rate has doubled directly.
Second, use vertical domain data to train AI. For example, if you do maternal and infant content, you will specifically feed AI with parenting knowledge, and the sounds and speed it generates will be more accurate. Last week, I helped a parenting blogger do this, and the audience commented that "it feels like my mother is talking."
Third, add real interaction. AI can generate Q & A videos, but you can respond in real people at key points. I tested it last week, and the interaction rate of videos made in this way was 30% higher than that generated by pure AI.
Many tools are now testing new functions internally. For example, Runway's Gen-2 can already generate 3D animations based on text. We tried it last week and typed in "drink coffee on rainy days", and it really generated a coffee scene with water drop effects.
The future trend is clear. AI will increasingly understand industry details. For example, for e-commerce, AI can automatically generate subtitles for product selling points; for teaching, it can identify professional terms and optimize pronunciation. Last week, I helped an English teacher test, and AI was able to automatically mark out the linking part.
But don't rely too much on AI. Last week, there was a customer that relied entirely on AI, but the audience found that his voice was too fake. Later, we changed to AI+ real person mashup, and the feedback was much better.
A little advice. Many tools are now available in trial versions. For example, coze's weekly free credit is enough for you to test several accounts. Last week, I helped my client plan this way, and I developed two popular accounts in one month.
Video production tools are changing, but what viewers want to see is still reality. AI can help you save time, but you have to come up with the ideas yourself. Last week, I tested a new gameplay. AI generated the beginning, real people told the story in the middle, and AI summarized the end. It turned out that this mix and match was the most eye-catching. "
