http://j.people.com.cn/n3/2024/0111/c95952-20120738.html
A free feature released by AliCloud’s “Tongyi Thousand Questions” app allows users to enter the experience page by typing in words such as “Tongyi Dance King” or “Zenmin Dance King” in the app. The original image’s facial expressions, body shape, fashion, background, and other characteristics will be reflected in the video. As a first step, “Tongyi Thousand Questions” offers users 12 popular dance templates, including “Subject Three,” “Mongolian Dance,” “Paddle Step,” and “Melbourne Shuffle.
The algorithm behind this feature is “Animate Anyone,” a video generation model developed by the Ali Tsongyi Laboratory. This research became popular on international SNSs such as X (formerly Twitter) and YouTube, with more than 100 million views of related videos and more than 10,000 stars on Github, a community for developers, making it the most popular base model in the recent base model field It has become one of the most popular algorithms in the field of foundational models.