Zoom wants to turn you into an AI-animated photorealistic avatar, but not until sometime next year.
The next feature, announced today at Zoom's annual development conference, turns a video clip that users record of themselves into a digital clone, complete with a head, upper arms, and shoulders. Users enter a script of what they want the digital double to say, and Zoom generates audio that syncs with the avatar's lip movements.
Smita Hashim, Zoom's chief product officer, told TechCrunch that the custom avatars are designed to help people chat “asynchronously” with colleagues in a “faster, more productive” way. Ta.
“Avatar saves users valuable time and effort in recording clips, allowing them to scale their video creation,” Hashim said.
However, it can also pose a risk of deepfakes.
A number of companies have developed AI techniques that digitally “clone” a person's face and combine the clone with a synthetic voice that sounds reasonably natural. For example, Tavus helps brands create virtual personas for personalized video ads, and Microsoft last year launched a service that can generate convincing digital stand-ins for people.
However, many of these tools implement certain strict safeguards to prevent abuse. Tavus requires verbal consent, and Microsoft requires customers to obtain written permission and consent from the avatar talent they are performing with.
Zoom was a little more vague about its safety measures.
Hashim noted that Zoom's usage policies prohibit abuse, and said the company has built in “many safeguards” in its custom avatar feature, including “advanced authentication” and watermarks.
“We will continue to review and add safety measures as necessary,” Hashim said. “We employ technology (…) to (…) make it clear that clips were generated using Avatars and (…) ensure the integrity of content generated by Avatars. .”
Mockup of Zoom's custom avatar feature. Image credit: Zoom
Zoom's digital images align with CEO Eric Yuan's broader vision to one day create an AI that can speak in Zoom meetings, respond to emails, and answer phone calls.
But these lookalikes come at a time when deepfakes are spreading like wildfire across social media, making it difficult to distinguish truth from disinformation.
So far this year, deepfakes featuring President Joe Biden, Taylor Swift and Vice President Kamala Harris have racked up millions of views and re-shares. Most recently, Hurricane Helen flooded the web with fake AI-generated images depicting destruction and human suffering.
Deepfakes are also used to target individuals, for example by impersonating a loved one. According to the FTC, losses related to identity fraud exceeded $1 billion last year.
How exactly will Zoom prevent scammers from using its tools to generate videos of people saying things they didn't say for malicious purposes? It's not yet clear there is no. The mockup provided by the company shows a watermark in the top right corner of the custom avatar video. However, such watermarks can be easily cropped by screen recording tools.
We expect to have further information in the first half of 2025. At that time, Zoom plans to release custom avatars for its asynchronous video tool Zoom Clips as part of a premium add-on that costs $12 per user per month.
Regardless of what steps Zoom ultimately takes, or not, regulatory efforts to stem the flood of deepfakes continue.
The United States has no federal law criminalizing deepfakes, and more than 10 states have enacted laws banning AI-based impersonation. California's law, which is currently stalled, would be the first to give judges the power to order deepfake posters to take them down. Failure to do so may result in fines.