no.571
Author: Tech Boomerang.
Proofreading: Cheng Cheng Editor: Zhizhi.
The field of AI is really getting more and more lively, and people are overwhelmed. Recently, Alibaba released a generative AI model called emo, which sounds a little strange, but it is really something to play. All you need is one ** and a piece of audio to generate a speaking and singing type**. The operation is relatively simple, and there is no limit to the length of time. According to the case, the emo generated**, the dynamics and expressions are very realistic, so that many users have a lot of fun.
At the end of February, Alibaba released its latest AI achievement, the eMote Portrait Alive (EMO) framework. This is an innovative framework developed by the Alibaba Institute of Intelligent Computing. Unlike other AI frameworks, eMO's main function is to generate expressive portraits through audio.
The biggest difference lies in its synthesis method. At present, the industry's mainstream AI synthesis technology is driven by 3D models or facial markers.
The synthesis of emo is more complex. From audio synthesis to audio compositing, a number of complex techniques are used. For example, audio-driven generation techniques that compose a character's head directly from a given image and audio clip.
One of the more important technologies is the stability control mechanism. A lot of AI synthesizers can be seen at a glance, because the facial expressions are stiff and don't match the verbal actions at all. The emo introduces a speed controller and a facial area controller, which can control the facial micro-expressions and make it more expressive. **Made by netizens**, the characters can also have delicate changes in facial expressions according to the emotional changes of the song when singing, which is very expressive.
I browsed the AI** made by other netizens through emo, and it was very fun. Some netizens resurrected their idols, and some netizens resurrected the historical figures in the textbooks, there are a lot of funny**, and everyone has a lot of fun. Netizens joked that with emo, there will be no more emo.
However, the emo page is all in English**, and after using the web page translation, the language is weird. If you have better English, you can try to play it. The web entrance can be opened by clicking here or by copying
Of course, this emo is just a literati simulacrum**, but it is made more smooth and realistic, logically, it is still very different from Sora.
In the past few years, Ali has spent a lot of money and put a lot of effort into AI. Just look at the introduction of the team behind the project on the emo webpage, and you can see that Ali has burned a lot of money on emo.
As early as last year, Wu Yongming, CEO of Alibaba, proposed an "AI-driven, public cloud-first" strategy, with the goal of providing efficient and stable AI infrastructure for various industries. Ali launched a number of AI products last year, such as the language model Tongyi Qianwen, the AI shopping guide for C-end customers, the painting model Tongyi Wanxiang, the AI photo tool Miao Ya Camera, and so on.
At present, the AI products launched by major Internet companies are really overwhelming, and AI has become ubiquitous on the Internet.
In addition, tools such as Jianying under Bytes have also launched many AI functions.
However, the extensive use of Wensheng's AI tools has also raised a lot of concerns. Mainly, a well-timed issue, the issue of image rights and fraud.
Users only need to upload ** and audio to generate smooth **, which can easily involve portrait rights issues. This issue is not uncommon, and there have even been many precedents. In addition, many netizens are worried about fraud. With the improvement of AI generation technology, should the technology of AI identification also be improved?
Finally, I would like to remind you that emo is not open source at present, so just play to prevent being deceived.