A creator stands before a phone QR scan as a luminous synthetic double forms inside a video-editing interface, with visible provenance markings and platform tiles for Flow, Gemini and YouTube.📷 AI-generated image / TECH&SPACE
- ★Google is adding avatars to Flow, Gemini and YouTube for AI video using a user’s likeness and voice.
- ★Setup uses a QR code, a spoken number sequence and head movement to capture several angles.
- ★Videos generated with the Omni model receive Google’s SynthID watermark, but the distribution risk remains social.
Google’s new tool inside Flow lands exactly where creative automation and identity safety collide. According to Wired’s report, Google has added an avatar feature to its AI video suite: a user scans a QR code on a phone, records themselves saying a string of numbers while moving their head, and the system can then place that digital likeness into generated videos.
This is not just another face filter. The point is that a creator does not need to stand in front of a camera every time they want to appear in a clip. Google’s framing, from the supplied context, is blunt: the feature is for creators who want to bring themselves into content without having to shoot themselves. Another quoted description goes further, saying the system can capture voice and visual identity from several angles and reproduce them with high fidelity.
A new avatar feature in Flow, Gemini and YouTube lowers the bar for videos where users look and sound like themselves, with SynthID watermarking as Google’s main safety layer.
A close, procedural view of identity capture: phone camera framing a face, number prompts, head-angle guides and a generated avatar timeline assembling frame by frame.📷 AI-generated image / TECH&SPACE
The technical engine is the new Omni Flash model, which in this context succeeds Veo and is described as improving detail and consistency, especially for user-generated avatars. The feature is not being kept inside one isolated app: the supplied context says it is available through Flow, Gemini and YouTube. That means Google is not merely testing a lab demo. It is building a path for synthetic presence to become a normal publishing workflow.
Editorially, the key word here is not “avatar.” It is “consent.” When a person deliberately records their own profile, the tool can be a legitimate production shortcut: a faster intro, a localized version, a reaction clip without setting up lights and a camera again. But the same infrastructure also trains audiences to accept videos where the line between a recorded statement and a generated performance is harder to see. If the face is persuasive, the voice is familiar and the format looks like ordinary social video, verification pressure shifts onto viewers, platforms and distribution rules.
Google is therefore leaning on SynthID watermarking for videos generated with the Omni model, including avatar videos. That is a necessary layer, but not a magic lock. A watermark helps establish provenance where systems can read it and where platforms decide to act. It does not automatically solve cropped clips, reposts, screenshots, political context or social harm that spreads before verification catches up.
That is why this story is bigger than one Google feature. Flow was launched last year under the experimental Labs umbrella as Google’s product line for creative work, and it is now moving into identity-sensitive production. If the avatar tool proves useful, the question will no longer be whether users can make deepfakes of themselves. The harder question will be who gets to confirm that such a video is really theirs, how consent can be withdrawn, and how quickly platforms can show that a synthetic person is synthetic after all.
