A Los Angeles startup has launched an impressive video tool that can show the person being filmed speaking a different language than the one they initially speak. After ChatGPT and Midjourney, this new artificial intelligence is also controversial.
You may have come across a video on social media of Lionnel Messi at a press conference speaking perfect English, or a video of Elon Musk or Donald Trump speaking French, or perhaps one of General de Gaulle speaking Italian or Portuguese . This is indeed a new artificial intelligence tool.
In the video above, Jon Finger, a screenwriter, tests the tool, simply called Video Translate, while sitting outside his house. He explains what he must do: film himself briefly and send the file so that the system processes his face and voice. He then gives the same explanation, but this time in French, with a slight Quebecois accent, then in German. His lips move as if he were really speaking French or German. The result is spectacular even if, when listening, there is a somewhat robotic echo in his translated voice.
Video Translate is still in beta but can already translate to English, French, German, Spanish, Italian, Polish, Portuguese and Hindi. Short videos are free but longer versions require a fee.
>> Faced with the global threat of false satellite images, geographic deepfakes, a Toulouse start-up finds a solution
A start-up specializing in the creation of video avatars
This invention is the creation of a start-up called HeyGen, based in Los Angeles. Its founder, Joshua Xu, was trained by Carnegie Mellon University in Pittsburgh, one of the best engineering schools in the world. He then worked for Snapchat from 2014 to 2020.
HeyGen was initially called Movio. Before Video Translate, its novelty, the start-up offered to create your own avatar. The advantage is that it is possible to control it, to make it say any text by typing it into the computer. The avatar is a hyper-realistic copy of his own face and his own voice, reproduced using a two-minute video that can be shot simply with a smartphone, no need for a studio or a high-definition camera. Beyond the technical prowess, it is also the ease of the process that impresses.
HeyGen also offers impersonal avatars, with a generic face and voice, which can be used to transmit a message, on a company’s website for example, without paying an actor and without a camera.
Concerns about the potential of this technology
Dishonest individuals will inevitably think of using it for criminal activities or disinformation. Changpeng Zhao, the founder of the cryptocurrency platform Binance, has expressed public concern because video identification is used by his company for financial exchanges.
Joshua Xu, the founder of HeyGen, argues above all that his tool can break down the language barrier for millions of people. Online video content would become easily accessible to the 90% of people who do not speak English, for example.
>> Strike in Hollywood: with the rise of artificial intelligence, stuntmen see their jobs threatened
But we can also imagine that, in the not-so-distant future, people could ask ChatGPT to write a text on any theme. They would then slide the text written by the artificial intelligence into the HeyGen tool and ask it to translate it into several languages using one of the generic avatars offered by the site. They would have produced content at a lower cost – no scriptwriter, no presenter, no cameraman, no equipment, no translator – with a minimum of effort.