YouTube adds AI avatars

YouTube is starting to let creators make a digital stand-in that looks and sounds like them, then drop that stand-in into Shorts without filming a new clip each time. The rollout began this week, and YouTube says it is gradual rather than instant for everyone. (support.google.com) (engadget.com) The setup starts with what YouTube calls a live selfie, which means recording your face from multiple angles and reading prompts out loud so the system can capture both your appearance and your voice. You only have to do that once unless you want to retake it to match a new haircut, glasses, or other change. (support.google.com) (9to5google.com) After that, the avatar can do two jobs. It can be inserted into an existing eligible Short through Remix and Reimagine, or it can star in a brand-new generated scene built from a text prompt. (support.google.com) (theverge.com) The generated clips are short on purpose. Reports on the launch say each prompt-made segment is about eight seconds long, which fits the quick-cut style YouTube has been pushing across Shorts. (9to5google.com) (engadget.com) This did not appear out of nowhere. YouTube spent 2025 adding image-to-video tools, artificial intelligence backgrounds, and Google DeepMind’s Veo video model to Shorts, so the new avatar is really the next layer on top of an app that was already moving from camera-first to prompt-first creation. (blog.youtube 1) (blog.youtube 2) The company is limiting who gets it first. YouTube’s help page says you must be at least 18 and be the owner of an existing YouTube channel, and the launch is global outside Europe rather than worldwide. (support.google.com) (engadget.com) That Europe exception is a clue about what this product really is. A tool that copies a person’s face and voice touches privacy law, biometric data rules, and artificial intelligence disclosure rules all at once, which is exactly where European regulators have been more aggressive than the United States. (engadget.com) (support.google.com) YouTube says the selfie video and voice are used to create the avatar, that other people cannot use your avatar to make original Shorts, and that you can delete the avatar later. The company’s help materials also say unused avatars are automatically deleted after three years, although videos you already published stay up until you remove those separately. (support.google.com) (letsdatascience.com) The platform is also leaning on labels instead of pretending viewers will somehow just know. YouTube has said its generative tools use SynthID watermarks and clear labels, and outside reporting on this avatar launch says the clips also carry Coalition for Content Provenance and Authenticity markers. (blog.youtube) (letsdatascience.com) For creators, the obvious use is low-friction filler work: a recipe intro in Spanish, a quick explainer recorded at midnight without setting up lights, or a travel update made while the real person is off camera. For viewers, the harder question is whether a face that used to mean “this person really said this here and now” will still mean that on Shorts six months from now. (theverge.com) (engadget.com) That tension is why this launch will probably land differently by category. A movie reviewer or language tutor can use an avatar as a timesaver, but a local news host, political commentator, or financial advice channel is stepping into a trust problem the second viewers have to wonder whether the person on screen was ever actually on camera. (theverge.com) (support.google.com)