YouTube Shorts debuts new AI video creation features

YouTube will now let creators create Shorts with AI avatars of themselves. The company had previously shared that it would be rolling it out at the start of the year, and now the functionality is now live on YouTube. Once properly set up, the AI-generated Shorts should look and sound like the creator.

The setup process involves capturing a “live selfie,” where users record their face and voice by reading a series of prompts. That data is then used to build a photorealistic avatar that can appear in Shorts. Individual AI-generated clips can run up to eight seconds, but creators can stitch multiple clips together. The avatar only needs to be created once, though it can be updated later if needed.

You'll only be able to use your own avatars in videos, and you can't use anyone else's avatar to create shorts, and vice versa. YouTube Avatars are automatically deleted from the company's servers after three years of disuse.

Once set up, you'll be able to create videos using text prompts, powered by Google's Veo text-to-video technology.

YouTube is rolling out this feature globally today, April 09, with the exception of Europe on the main YouTube app and YouTube Create, and it's available for users 18 and older with existing YouTube channels.

"You'll be able to create a Short using your own likeness, produce games with a simple text prompt, and experiment with music. Throughout this evolution, AI will remain a tool for expression, not a replacement," YouTube CEO Neal Mohan said earlier this year.

Deepfakes in your pocket

Deepfakes (or AI-generated videos of real people) are a hot-button topic in generative AI space, as people often fall for hoax videos — or conversely mistake real videos for AI-generated ones. Giving users the power to easily generate deepfakes of themselves is something that may raise an eyebrow among spaces where AI use has frequently been criticised for its dangers in an increasingly artificial world.

YouTube is aware of this, and the company says that it'll clearly mark AI-generated content out as AI-generated on its platform.

"It’s becoming harder to detect what’s real and what’s AI-generated. This is particularly critical when it comes to deepfakes. We clearly label content created by YouTube’s AI products, and creators must disclose when they've created realistic altered or synthetic content," Mohan said. 

Separate to this, YouTube has also rolled out an AI deepfake detection tool for politicians and journalists. The company says its committed to reinforcing trust on its platform. Still, as these tools become easier to use and more accessible, the broader challenge, distinguishing reality from fabrication, isn’t going away anytime soon.