Chinese tech company ByteDance, the owner of social media platform TikTok, has introduced a new service called OmniHuman-1, which allows users to generate videos from images using artificial intelligence (AI), according to TechCrunch.
Analysts from the publication noted that the videos generated by OmniHuman-1 are currently among the most realistic available.
«AI for deepfakes (a term used for videos or images generated by AI – Ed.) has become a mainstream consumer product. There are numerous applications capable of inserting a person into a photograph or making someone appear to say something they never actually said. However, most ‘deepfakes’ – particularly videos – fail to overcome the ‘uncanny valley’ effect,» the article states.
The uncanny valley effect refers to the discomfort or aversion people feel when they see an object that acts like a human but still appears artificial.
Judging by the sample videos created using OmniHuman-1 and selected by ByteDance, the company appears to have overcome this limitation. If this is the case, the ethical concerns surrounding the use of deepfakes are likely to become even more pressing.
