Zoom is preparing to let users create photorealistic AI versions of themselves for work videos. The feature is aimed at making recorded communication faster, but it arrives at a moment when synthetic likenesses are already creating serious trust and safety problems online.
How Zoom's custom AI avatars are supposed to work
The upcoming tool was announced at Zoom's annual dev conference. It will take a video clip that a user records of themselves and turn it into a digital clone with a head, upper arms and shoulders.
Once the avatar exists, the user will be able to type a script. Zoom will then generate audio for that script and sync the speech with the avatar's lip movements. In practice, that means a person could produce a video message without sitting down to record the full clip on camera.
Zoom is positioning the feature around asynchronous communication. Smita Hashim, Zoom's chief product officer, told TechCrunch that the goal is to help people communicate with colleagues "asynchronously" in a "faster, more productive" way.
Hashim also said, "Avatars save users precious time and effort recording clips, and enable them to scale video creation." That framing matters: this is not being presented only as a novelty. It is being tied directly to the everyday workflow of sending video updates, creating clips and communicating across teams.
Why the same convenience creates risk
The same qualities that make custom AI avatars useful for workplace video can also make them sensitive. A photorealistic digital double that can say typed words is, by design, close to the kind of technology used in convincing deepfakes.
Other companies have already built systems that digitally clone a person's face and pair that clone with reasonably-natural-sounding synthetic speech. Tavus, for instance, helps brands create virtual personas for personalized video ads. Microsoft last year launched a service that can generate convincing digital stand-ins for a person.
The source article points out that some companies have added strict safeguards around this kind of technology. Tavus requires verbal consent statements. Microsoft mandates that its customers obtain written permission and consent from any featured avatar talent.
Zoom's described safety approach is less specific. Hashim pointed to Zoom's usage policies prohibiting misuse and said the company is building "numerous safeguards" into the custom avatar feature. Those measures include "advanced authentication" and watermarking.
Hashim also said, "We will continue to review and add safeguards as needed in the future." She added that Zoom uses technology to make it clear when a clip is generated with an avatar and to help ensure the integrity of avatar-generated content.
The watermark question
One visible safeguard appears in a company mock-up: a watermark in the upper-right-hand corner of a custom avatar video. That could help viewers understand that a clip was generated rather than recorded in the ordinary way.
But the source article raises a practical concern. Watermarks like these can be cropped out by screen-recording tools. That does not mean watermarking is useless, but it does show why a visible label alone may not answer every misuse scenario.
The central question is straightforward: how will Zoom stop scammers or other bad actors from making videos of people saying things they did not say? Based on the information provided, the complete answer is not yet clear.
That uncertainty is important because this feature is connected to a wider product direction at Zoom. The digital likenesses fit with CEO Eric Yuan's broader vision of AIs that could one day speak in Zoom meetings for users, answer emails and take phone calls.
Deepfakes are already a mainstream problem
The timing makes the debate more urgent. Deepfakes are spreading across social media and making it harder to separate truth from disinformation.
So far this year, deepfakes involving President Joe Biden, Taylor Swift and Vice President Kamala Harris have drawn millions of views and reshares. More recently, fake generative AI images of destruction and human suffering appeared across the web after Hurricane Helene.
The risk is not limited to public figures or major events. Deepfakes have also been used against individuals, including through impersonation of loved ones. Losses linked to impersonation scams topped $1 billion last year, according to the FTC.
For a workplace tool, that context changes how the feature may be judged. Users may value speed, but organizations also have to think about verification, consent, trust and what happens when generated media leaves its original platform.
What comes next for Zoom Clips
Zoom plans to release custom avatars for Zoom Clips in the first half of 2025. Zoom Clips is the company's asynchronous video tool, and the avatar feature is expected to be part of a $12 per-user, per-month premium add-on.
By then, the company may provide more detail about how authentication, watermarking and other safeguards will work in practice. Those details will matter because the usefulness of AI avatars depends on whether people can understand when they are seeing generated content and whether misuse can be meaningfully limited.
Regulators are also moving, though the source article notes that there is no federal U.S. law criminalizing deepfakes. More than 10 states have enacted statutes against AI-aided impersonation. California's law, currently stalled, would be the first to empower judges to order posters of deepfakes to take them down or potentially face monetary penalties.
Zoom's custom AI avatars could make routine video communication easier. They also place the company in the middle of a larger test for synthetic media: whether productivity tools can offer realistic digital likenesses without making impersonation easier to scale.