YouTube adds AI avatars
YouTube has started letting creators generate photorealistic AI avatars for Shorts from a selfie and voice scan, and those Shorts will carry visible labels that the clip used AI. This lowers the cost and time to prototype host-led or character-led short-form content, but it also raises questions about authenticity and how viewers—or acquirers—should read engagement metrics when synthetic clips scale quickly. (phandroid.com, es.wired.com)
YouTube is starting to let creators make Shorts without getting back on camera each time. A creator can record one “live selfie” and one voice sample, then ask YouTube to generate a photorealistic version of them for new clips. (support.google.com) The rollout is not for everyone at once. YouTube says the feature is rolling out gradually, and the account has to belong to a channel owner who is at least 18 years old. (support.google.com) This did not come out of nowhere on April 9, 2026. YouTube chief executive Neal Mohan previewed the plan on January 21, 2026, when he wrote that creators would be able to “create a Short using your own likeness” this year. (blog.youtube) YouTube has been building toward this through its Shorts artificial intelligence tools for months. In 2025 it added image-to-video and other Veo-powered generation features, and it said those creations would carry clear labels and SynthID watermarks. (blog.youtube) The new step is that the synthetic person is no longer a stock character or a moving background. The avatar is built from your own face and voice, so YouTube is turning the creator into a reusable asset inside the Shorts editor. (support.google.com, phandroid.com) YouTube says the capture process is a secure “live selfie,” and it says only the account owner can use that avatar to create original videos. The company also says the face and voice capture are used to create the avatar, not opened up for other users to make clips with your likeness. (support.google.com, 9to5google.com) Viewers are supposed to know when this is happening. YouTube’s help pages and product coverage say artificial intelligence-generated Shorts from these tools will include visible disclosure labels, and YouTube has already tied its Shorts artificial intelligence outputs to provenance systems like SynthID. (blog.youtube, letsdatascience.com) For creators, the obvious use is speed. Instead of setting up lights, recording retakes, and editing around mistakes, they can prompt short host segments on demand, which makes testing a new character, language, script angle, or sponsor read much cheaper. (phandroid.com, es.wired.com) For everyone else, the awkward part is what a view now means. If one person can spin up many polished host-led clips from a single scan, then engagement starts measuring prompt quality and distribution as much as filming skill or on-camera labor. (es.wired.com, engadget.com) That matters most in corners of YouTube where buyers care about a creator’s “face time.” A media company, talent manager, or brand looking at a fast-growing Shorts account now has to separate audience demand for a person from audience demand for a synthetic version of that person that can publish at machine speed. (es.wired.com, blog.youtube) YouTube’s bet is that labels and account-level controls will be enough to keep this in the “tool” bucket instead of the “deepfake” bucket. The next test is whether viewers treat an artificial intelligence host as a shortcut, a gimmick, or just another normal part of the Shorts feed. (support.google.com, blog.youtube)
