The logo of the American multinational Microsoft is visible at the Mobile World Congress (MWC), the largest annual gathering of the telecommunications industry, in Barcelona, March 2, 2023. (AFP/Archives/Josep LAGO)
Microsoft researchers have developed artificial intelligence (AI) technology that will turn a photo of a face and a voice sample into an ultra-realistic video of a “speaking face”, according to a published document by the IT giant this week.
“The objective is not to create content intended to mislead or deceive,” the company specifies, but it recognizes that like “other content generation techniques, it could be used improperly to impersonate a human being.
The rapid growth of generative AI, which makes it possible to easily produce all kinds of content (texts, images, sounds, etc.) of astonishing quality, raises many concerns, particularly in terms of exploitation for the purposes of fraud and misinformation.
“Our research focuses on (…) virtual avatars, with a view to positive applications,” assures Microsoft. “We are opposed to any behavior aimed at creating misleading or damaging content from real people.”
The IT firm, the main investor in OpenAI (ChatGPT), therefore does not plan to make the new tool available or provide technical information “until we are certain that the technology will be used responsibly and in accordance with regulations in force”.
Called VASA-1, the computer program takes a simple portrait photo and an audio file and converts them into a video where the face animates and speaks in a hyper-realistic way.
In the examples posted online – previews made from images of virtual characters – the movement of the lips is synchronized, the facial features convincing and the movements of the head, a little jerky but still impressive.
Other companies are working on this technology, such as Runway, one of the specialists in generative AI for video. Google researchers have also created an AI model, Vlogger, that can generate realistic videos of talking heads.
According to Microsoft, the benefits of developing this technology “such as increasing educational equity, improving accessibility for people with communication difficulties, providing therapeutic support to people who need it, among others” justify carrying out this research.
Many authorities are considering regulating generative AI. The European Union agreed in December on unprecedented legislation to regulate AI, seeking to encourage innovation while limiting possible abuses.
© 2024 AFP
Did you like this article ? Share it with your friends using the buttons below.
