Zoom has revealed plans to launch photorealistic, AI-animated avatars that will turn recorded video snippets into digital portraits of users. This feature was unveiled at Zoom’s annual developer conference and is scheduled to launch the following year. These head, shoulders, and upper arm replicas of the user are capable of speaking pre-written lines with audio produced by AI that matches the avatar’s lip movements.
Zoom’s chief product officer, Smita Hashim, claims that the avatars are intended to speed up and simplify asynchronous communication by removing the requirement for users to record entire videos in order to share messages. Hashim emphasised the effectiveness of avatars, saying that they enable users to create video content more quickly and efficiently.
But there are worries about the misuse potential, especially in light of the growing popularity of deepfakes. Concerns have previously been raised by AI systems that can clone faces and produce synthetic speech. Strict measures are taken to prevent misuse of similar features, such as requesting written or verbal agreement before making digital reproductions, by corporations like Microsoft and Tavus. Although Zoom emphasizes its usage policies, it has not yet provided a detailed description of the steps it would take to stop abuse. In order to recognize avatar-generated content and guarantee its integrity, the business has stated that it is planning to implement “advanced authentication” and watermarking.
There is a growing risk of fraud and misinformation as deepfakes spread. Deepfake videos of celebrities like Taylor Swift and President Joe Biden have gone viral just this year, and impersonation scams have cost victims over $1 billion in losses. While Zoom has alluded to precautions, it is unclear how the company can shield customers from unethical use of its avatar technology. As the feature approaches its 2025 launch, more information is anticipated. The Zoom Clips premium add-on will cost $12 per user every month. State and federal initiatives to control deepfakes are still ongoing in the interim.