Google has updated its AI video tool Flow with a new feature that allows users to create AI-generated videos of themselves using a personal avatar, the company announced in May 2026.
The feature works by having users scan a QR code within their Flow account settings on their phone, then record themselves reciting a string of numbers and rotating their head to capture multiple angles. The resulting avatar can then be inserted into AI-generated video scenes, with the ability to adjust details like background settings and clothing while preserving the likeness and voice of the user.
The update also introduces Omni Flash, a new video-generation model replacing Veo. According to Google, Omni Flash improves character consistency across successive video generations — a noted weakness in earlier versions of Flow, where characters could visually distort between clips. Every video generated using the Omni model, including avatar videos, carries Google’s SynthID watermark.
Google’s approach currently limits avatar generation to users depicting themselves only, not other people. This differs from OpenAI’s Sora app, which allowed users to generate videos of other users depending on account settings. OpenAI shut down the Sora app after less than seven months.
This is not Google’s first move in this space. Last month, YouTube Shorts added a limited option for users to create similar AI avatars for use in clips on that platform. Meta has also introduced a related feature that translates Instagram Reels into different languages and adjusts creators’ lip movements to match.
The expansion of self-deepfake tools across major platforms may streamline content creation for some users, but the source notes that generative AI video is increasingly dividing audiences who view such content as inauthentic — provided they recognize it as AI-generated at all.
Source: WIRED