From the birth of ChatGPT to the beginning of 2024, just over a year has passed.
The exploration of generative AI by major manufacturers has entered the next stage.
In the beginning, it was a carnival of a handful of geek users.
Now, hundreds of millions of ordinary users can enjoy the benefits of AIGC.
For example, Xiaomi's Xiaoai, vivo's Blue Heart V, Huawei's Xiaoyi, and OPPO's Xiaobu all have the ability to connect with contextual conversations.
Let the voice assistant become a personal intelligent assistant, this is the charm of the AI large model.
And get past the hardest beginning.
This was followed by a massive explosion of AIGC's multimodal capabilities.
In other words, today's AI can not only generate images from text, but also turn images into videos, or even text into videos...
No matter how strong something is, it still has to be used.
What will the open-minded, good-speaking netizens do with generative AI?
At first, I thought that everyone would use it to help them learn Xi, create copywriting, and draw inspiration.
But as it turned out, it was I who was thinking too complicated.
In this Internet where fun is king, the first purpose of netizens is to live.
For example, a while ago, I swiped the whole network's subject three dance.
The silky and magical dance steps made many netizens dance spontaneously.
But the question arises – if I want my friends to jump too, how do I do that?
换做前。
Brother Ji will advise you to invite a friend to a set meal on Crazy Thursday and make him a little ugly.
But now, times have changed.
Ali's Tongyi Qianwen App has recently released a [National Dance King] function.
All you have to do is send a full-body photo of your limbs, body, and head to the app, and then select a dance template.
In about ten minutes, you will receive the finished video.
As the saying goes, the lower the threshold for getting started, the more ruthless netizens will be.
Nowadays, there have been videos of various celebrities, entrepreneurs and even ancients jumping "subject three" on the Internet.
For example, this old man who wears suspenders and has a full sense of dance and rhythm.
I don't say it, everyone should know who it is.
The king of Hercules, who has the effect of "difficult to stretch", has also become the king of the three dances of the subject with the blessing of AIGC.
Even the football king Messi was used as a living material by netizens and generated a subject three videos.
Don't say it, this smart dance, this swaying rhythm, is not even inferior to the original.
However, watching others live a hundred times is not as happy as getting started once.
Brother Ji will talk about it in detail, how to use it, and what details need to be paid attention to in the process.
Tongyi Qianwen's App, everyone should be able to search for it in their own mobile phone app store, so I won't put a link in the article.
Let's open the App directly and see these feature recommendations.
There is no need to hesitate, click on "A photo to dance", and you will jump to the [Dance King for All] section.
There are more than a dozen dance templates built into it, and it is certainly not rich to say that it is rich.
But it still has a hand in the heat.
For example, subject 3, which has been very popular recently, and the blissful dance that has been popular in the past, all have corresponding templates.
Since he has come, Brother Ji must choose "subject three".
The next step is the much-anticipated custom image link.
Here we can find a full-body photo.
Whether it's a two-dimensional anime or a three-dimensional live-action, it supports generation.
Most importantly, the posture should not be too complicated, the light should not be too dim, and the hands should not hold things.
It can only be said that there are still many restrictions on the conversion of static images to AI videos.
The regular portrait full-body picture has been tried by many netizens.
Brother Ji didn't take the usual path this time, and directly took the teddy bear to do a test.
When the picture is uploaded, we can spread our hands and wait for the results to come out.
Because the image has to be processed in the cloud, it usually takes about 15 minutes to wait.
Compared to pure local GPU processing, the time is certainly not fast.
However, similar AI services either have too long a queue or cannot be accessed directly.
At this time, it is very easy to use...
As for the finished effect, it can be seen that the teddy bear is jumping subject three, but the details on the body and hands are still a bit strange.
Considering that there is only a little more than ten minutes of training time, it is okay to use it for fun.
Another Hokage man, Sasuke.
Brother Ji deliberately found a relatively clear full-body photo.
The face is clear, the limbs are distinct, and the material can't be blamed for not working.
And with the blessing of high-quality materials, Tongyi Qianwen did give a very good finished product.
There is no distortion in the facial features, and even the finger movements are clearly depicted.
Materials with a clear sense of line, such as character animation, are simply not as difficult as today's AI.
As for what you said, the quality of this video can still be improved.
I think so, after all, generative AI has one obvious characteristic:
The more footage and the longer the training time, the better the quality of the finished film. But as a function for the majority of ordinary users, it can't make us wait for more than half an hour.
Therefore, such a production time and quality should be a more balanced configuration.
It is not only suitable for friends to prank each other, but also can be used for self-entertainment.
In addition to Ali's Tongyi Qianwen.
ByteDance also has a similar image-to-video AI technology - MagicAnimate.
The actual effect is similar to that of Tongyi Qianwen, which can turn a static picture into a dynamic video.
Multimodal models such as Tusheng video and Wensheng video have been swept up by large manufacturers.
In the long run, a few more iterations of this kind of technology may be able to replace the painter and do some tedious redrawing work.
But the eighth brother, who likes to watch beautiful women on Douyin on weekdays, put forward different opinions at this time:
"In the future, beautiful bloggers can generate hot dance videos with a picture, so what else do I watch?"
There are even netizens who have a very bold idea in their hearts.
That's true.
If the poses are fully customizable, you might end up creating outrageous videos.
That's why apps and web pages that offer this type of experience only provide fixed pose templates.
Find the subject 3 template inside, and it's no problem to have fun on the hour.
If you really have any bold ideas, I advise you not to think about it.