Is it a human speaking behind the camera or an AI clone? A stunning innovation from an Nvidia-backed unicorn startup is making it almost impossible to tell the difference.
AI startup Synthesia, which reached unicorn status with a billion-dollar valuation last year, released a new technology called Expressive Avatars on Thursday; the world’s first digital AI clones capable of generating human facial expressions and the correct tone of voice from written requests.
The technology starts with an AI avatar that can be customized to reflect real faces.
Photo credit: Synthesia
The AI creates a digital copy of a person based on footage captured on their webcam or in a certified studio. It can also clone the person’s voice to integrate it into their digital image.
Anyone concerned about creating an AI avatar that takes on their face and voice can instead opt for one of the more than 160 pre-installed AI avatars that Synthesia has in its database.
Related: ‘This is a serious problem’: Mr. Beast criticizes AI deepfakes
Once a user creates or selects an AI avatar, all they have to do is write what they want to say with their digital self.
In a demo seen by CNBC, one user wrote: “I’m happy. I’m sad. “I’m frustrated.” and had the AI-generated digital clone read the text. The avatar conveyed facial expressions and tone of voice associated with happiness when saying the text prompt “I’m happy,” and changed his tone accordingly when he said, “I’m frustrated.” The tone matched the words.
Using an AI clone and a written prompt, a free user can create 36 minutes of personalized videos in more than 120 languages every year. Paid plans cost up to $67 per month for up to 360 minutes of video per year, or unlimited minutes of video for businesses that choose an enterprise plan.
Synthesia is a startup that large companies use behind the scenes. Zoom, Xerox, Microsoft and Reuters all use Synthesia’s programs internally. Synthesia CEO Victor Riparbelli told MIT Technology Review that 56% of the Fortune 100 are using the technology.
Synthesia markets the technology as a way to create expressive digital avatars for corporate training and presentations. For example, Zoom designers created sales training videos in Synthesia in 90% less time than it took humans to create the videos.
Related: JPMorgan says its AI cash flow software reduces human labor by nearly 90%
“Zoom subject matter experts no longer need to record themselves and have 15 to 20 hours each month to work on their actual work,” says the Synthesia website.
Still, the ability to create incredibly good deepfakes, or AI that clones and manipulates a human’s voice, likeness, or other aspects without their permission, can lead to abuse.
Last month, Tennessee became the first US state to pass a law protecting music industry professionals from deepfakes.