AI Model OmniHuman-1 Presented
Chinese company ByteDance, which owns the social network TikTok, has presented its own artificial intelligence (AI) model OmniHuman-1, which is capable of generating videos of people from photographs and audio, according to a statement from researchers who worked on the project.
The videos created by OmniHuman-1 look amazingly realistic. Just look at the video below, which shows a speaker giving a speech at the popular annual conference on the dissemination of unique ideas TED (technology, entertainment, design) that did not actually take place.
The OmniHuman-1 system only needs a single photo and a voice recording to generate such a video. The user can adjust elements such as aspect ratio and framing. Moreover, the AI is able to change body movements and gestures in real video footage. Of course, the results are not always 100% perfect. Some poses look a little unusual. For example, in this lecture by Einstein, completely generated by AI, the face looks almost perfect, but the hand movements are unnatural.
In December 2024, a report from the China Internet Network Information Center (CNNIC) reported that the number of users of generative artificial intelligence products in China had reached 230 million people.
