Google Gemini Omni avatar feature enables deepfake video creation
Google has rolled out a Gemini Omni feature that allows paid users to create hyperrealistic deepfake videos of themselves using selfies and voice data. The tool generates talking-head videos in minutes but raises concerns about potential misuse despite watermarking and age restrictions.
What happened
Google announced Gemini Omni at its I/O 2026 keynote, an AI model that generates videos from text, images, documents, and video clips. The company then rolled out a new Omni feature allowing users to generate hyperrealistic avatars of themselves using a couple of selfies and voice samples, creating talking-head videos in just a few minutes.
The feature is limited to paying Google AI Pro or AI Ultra subscribers who are at least 18 years old. All generated videos are watermarked with SynthID.
Reported impact
Testers reported the deepfake videos are so realistic that they raise concerns about potential misuse for creating false statements or scenarios. One tester noted creating a video of themselves admitting to a crime, demonstrating the tool's capacity to generate incriminating content.
Index analysis
The feature's capability to generate video content showing someone performing actions they did not perform or making statements they did not make presents risks for reputational harm and fraud. The ease of use—requiring only five minutes and a few selfies—lowers barriers to creating convincing fake videos.
Governance controls
Google restricts the feature to users 18 and older with paid accounts and marks all Omni videos with SynthID watermarks. However, concerns about deepfake video abuse remain unaddressed by these measures alone.
Classification
Sources and evidence
Sources not yet published
This record was researched and written by the Index from original reporting. Its sources are still being verified by an editor, and they are listed here once that check is complete. Until then the record carries no reviewer.